Spaces:
Sleeping
Sleeping
Commit ·
57ef382
1
Parent(s): c53d167
refactor + feat: KI-050b/053/054 — rename completeness + eval speed + tax questions
Browse filesCombines three logically-related changes into one push-clean commit
(rewrites the prior 67bdd3a + 866fb24 to drop accidentally-swept-in
binaries + runtime state that tripped HF Space's pre-receive hook).
KI-050b — data/ → 40-data/ rename completeness across docs:
• 40-data/ internal docs (README, corpus_urls.md, info_source_map,
illustrative_premiums.json, llm_health.json, profiles/README,
reviews/icici-lombard.json)
• kb/ — AUDIT_TRAIL, INDEX, all 218 kb/policies/*.md mirror sheets,
kb/research/corpus_acquisition.md, kb/reviews/* leaderboards
• tools/ — info_source_map.py, build_kb_mirror.py, full_pipeline.sh,
README, ~12 other scripts touching data/ paths in prose
• eval/info_source_map.json (2772 source-provenance entries) + README
entrypoint.sh preserved — its `/data/` paths are absolute HF Space
persistent-disk mounts, NOT references to the repo folder.
KI-053 — eval/run.py --no-extract flag. Skips the orchestrator's
free-form profile_extractor LLM call which wastes a NIM request per
gold-eval question (user_profile is always empty for these). Routes
through INSURANCE_BOT_SKIP_PROFILE_EXTRACTOR=1 env var. ~25% wall-time
saving on a rate-limited eval run. Production behaviour unchanged.
Combined with the existing --no-judge regex-grader flag:
• baseline: ~45 min
• --no-extract: ~32 min (-30%)
• --no-judge: ~15 min (-67%)
• --no-judge --no-extract: ~10 min (-78%)
KI-054 — 14 new tax-planning gold-QA questions (96 → 110 entries).
All `expected_refusal: true` because the corpus has zero Section 80D /
80DDB / 80U / 10(10D) / TDS / HRA / GST-rate content. Lets us measure
the refusal precision gap that drove the tax_planner archetype's 4.6×
refusal rate in the 100-persona audit.
Docs:
• CLAUDE.md — KI-045 (natural conv), KI-046 (refusal precision),
KI-047/050/051 (bucket reorg), KI-048/052 (admin), KI-049
(retrieval top-k); always-parallelize prime-directive line.
• 80-audit/ENTERPRISE_AUDIT.md — 12-row table extension covering
KI-036 through KI-052 with commit SHAs.
.gitignore — block 40-data/policies/, 40-data/sessions/, 80-audit/
full_*/, 80-audit/postfix_*/, MUST_FIX.md from future accidental
inclusion. These are runtime/large-binary paths that HF Space's
pre-receive hook rejects.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This view is limited to 50 files because it contains too many changes. See raw diff
- .gitignore +7 -0
- 40-data/README.md +4 -4
- 40-data/corpus_urls.md +28 -28
- 40-data/information_source_map.md +7 -7
- 40-data/llm_health.json +70 -50
- 40-data/llm_usage.jsonl +313 -0
- 40-data/premiums/illustrative_premiums.json +7 -7
- 40-data/profiles/README.md +3 -3
- 40-data/reviews/icici-lombard.json +1 -1
- 80-audit/ENTERPRISE_AUDIT.md +16 -2
- CLAUDE.md +27 -0
- README.md +4 -4
- backend/orchestrator.py +14 -2
- eval/gold_qa.json +155 -1
- eval/run.py +11 -0
- kb/AUDIT_TRAIL.md +8 -8
- kb/INDEX.md +1 -1
- kb/policies/acko__acko-health-ii__wordings.md +1 -1
- kb/policies/acko__acko-health-iii-platinum-lite__wordings.md +1 -1
- kb/policies/acko__acko-health-iii-platinum__wordings.md +1 -1
- kb/policies/acko__acko-health-iii__cis.md +1 -1
- kb/policies/acko__acko-personal-health__wordings.md +1 -1
- kb/policies/acko__arogya-sanjeevani__wordings.md +1 -1
- kb/policies/aditya-birla__activ-assure-diamond.md +1 -1
- kb/policies/aditya-birla__activ-assure-diamond__wordings.md +1 -1
- kb/policies/aditya-birla__activ-health-individual__wordings.md +1 -1
- kb/policies/aditya-birla__activ-health.md +1 -1
- kb/policies/aditya-birla__activ-one.md +1 -1
- kb/policies/aditya-birla__activ-one__brochure.md +1 -1
- kb/policies/aditya-birla__activ-secure-cancer-secure__brochure.md +1 -1
- kb/policies/aditya-birla__activ-secure-personal-accident-cancer-secure__wordings.md +1 -1
- kb/policies/aditya-birla__group-activ-health__wordings.md +1 -1
- kb/policies/bajaj-allianz__comprehensive-care-plan.md +1 -1
- kb/policies/bajaj-allianz__comprehensive-care-plan__wordings.md +1 -1
- kb/policies/bajaj-allianz__criti-care__wordings.md +1 -1
- kb/policies/bajaj-allianz__extra-care-plus__wordings.md +1 -1
- kb/policies/bajaj-allianz__global-health-care.md +1 -1
- kb/policies/bajaj-allianz__global-health-care__wordings.md +1 -1
- kb/policies/bajaj-allianz__group-health-guard-gold__wordings.md +1 -1
- kb/policies/bajaj-allianz__group-personal-accident__wordings.md +1 -1
- kb/policies/bajaj-allianz__health-guard-gold-individual__wordings.md +1 -1
- kb/policies/bajaj-allianz__health-guard-gold.md +1 -1
- kb/policies/bajaj-allianz__health-guard.md +1 -1
- kb/policies/bajaj-allianz__health-guard__wordings.md +1 -1
- kb/policies/bajaj-allianz__silver-health__cis.md +1 -1
- kb/policies/bajaj-allianz__tax-gain.md +1 -1
- kb/policies/bajaj-allianz__tax-gain__cis.md +1 -1
- kb/policies/care-health__care-advantage-add-ons-protect-plus-care-shield__brochure.md +1 -1
- kb/policies/care-health__care-advantage.md +1 -1
- kb/policies/care-health__care-advantage__brochure.md +1 -1
.gitignore
CHANGED
|
@@ -82,3 +82,10 @@ rag/_hf_dataset_backup/
|
|
| 82 |
# HF Space rejects binaries in the repo (use Xet/Releases instead).
|
| 83 |
# Regenerate locally any time via 'python tools/build_readme_pdf.py'.
|
| 84 |
README.pdf
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 82 |
# HF Space rejects binaries in the repo (use Xet/Releases instead).
|
| 83 |
# Regenerate locally any time via 'python tools/build_readme_pdf.py'.
|
| 84 |
README.pdf
|
| 85 |
+
|
| 86 |
+
# Runtime + heavy binary: never commit
|
| 87 |
+
40-data/policies/
|
| 88 |
+
40-data/sessions/
|
| 89 |
+
80-audit/full_*/
|
| 90 |
+
80-audit/postfix_*/
|
| 91 |
+
MUST_FIX.md
|
40-data/README.md
CHANGED
|
@@ -13,7 +13,7 @@ The structured policy schema and PDFs themselves live under `rag/`. This folder
|
|
| 13 |
| File | What it is | Owner |
|
| 14 |
| --- | --- | --- |
|
| 15 |
| `corpus_urls.md` | Discovery manifest — every PDF URL ingested into `rag/corpus/`. | discovery agent / `tools/check_link_rot.py` |
|
| 16 |
-
| `regulatory_urls.md` | IRDAI / regulatory PDF URLs. See [ADR-017](../docs/60-decisions/ADR-017-irdai-corpus-playwright-rescue.md). | discovery agent |
|
| 17 |
| `information_source_map.md` | Human-readable claim → URL → verdict map. Master audit doc for the Source Methodology directive. Mirror of `eval/info_source_map.json`. | `tools/info_source_map.py` |
|
| 18 |
| `llm_health.json` | Last per-provider health-probe snapshot (latency, success, last error). Powers the admin tab. | `backend/llm_health.py` |
|
| 19 |
| `llm_usage.jsonl` | Append-only per-call log: provider, model, tokens, latency, success. Aggregated in the admin tab. | `backend/main.py` |
|
|
@@ -26,7 +26,7 @@ The structured policy schema and PDFs themselves live under `rag/`. This folder
|
|
| 26 |
| `sessions/` | runtime | Per-session conversation state JSONs. Ephemeral — pruned periodically. Currently includes `anonymous.json` (no-name fallback). |
|
| 27 |
| `policy_facts/` | pre-computed | **256 curated JSONs**, one per policy variant. Each field carries `{value, unit?, source_pdf_path, source_quote}` provenance. The Indian-BFSI-audit-grade machine source; `kb/policies/*.md` are the human-readable mirror. See `_curation_report.md` for the three batches that built it. |
|
| 28 |
| `policies/` | pre-computed | Subfolder per insurer with PDFs / supplementary text used for one-off lookups outside the main ingest pipeline. |
|
| 29 |
-
| `premiums/` | pre-computed | `illustrative_premiums.json` — sample starting premiums pulled from PolicyBazaar / JoinDitto / Beshak + insurer rate cards (2026-05-13). Refreshed by `tools/refresh_premiums.py`. **Illustrative only** per [ADR-007](../docs/60-decisions/ADR-007-illustrative-pricing.md). |
|
| 30 |
| `reviews/` | pre-computed | One JSON per insurer with IRDAI claim-settlement metrics, complaints/10K, aggregator sentiment, news tone. Index + leaderboard in `reviews/INDEX.md`. Source: IRDAI Annual Report 2023-24. |
|
| 31 |
|
| 32 |
## Provenance + KPIs
|
|
@@ -41,5 +41,5 @@ The structured policy schema and PDFs themselves live under `rag/`. This folder
|
|
| 41 |
|
| 42 |
- [`kb/AUDIT_TRAIL.md`](../kb/AUDIT_TRAIL.md) — end-to-end lineage; `data/policy_facts/` is stage 8 output
|
| 43 |
- [`kb/INDEX.md`](../kb/INDEX.md) — policy index with completeness % per file
|
| 44 |
-
- [ADR-007](../docs/60-decisions/ADR-007-illustrative-pricing.md) — pricing is illustrative, never a real quote
|
| 45 |
-
- [ADR-009](../docs/60-decisions/ADR-009-19-insurer-comprehensive-schema.md) — 19-insurer scope + 48-field schema
|
|
|
|
| 13 |
| File | What it is | Owner |
|
| 14 |
| --- | --- | --- |
|
| 15 |
| `corpus_urls.md` | Discovery manifest — every PDF URL ingested into `rag/corpus/`. | discovery agent / `tools/check_link_rot.py` |
|
| 16 |
+
| `regulatory_urls.md` | IRDAI / regulatory PDF URLs. See [ADR-017](../70-docs/60-decisions/ADR-017-irdai-corpus-playwright-rescue.md). | discovery agent |
|
| 17 |
| `information_source_map.md` | Human-readable claim → URL → verdict map. Master audit doc for the Source Methodology directive. Mirror of `eval/info_source_map.json`. | `tools/info_source_map.py` |
|
| 18 |
| `llm_health.json` | Last per-provider health-probe snapshot (latency, success, last error). Powers the admin tab. | `backend/llm_health.py` |
|
| 19 |
| `llm_usage.jsonl` | Append-only per-call log: provider, model, tokens, latency, success. Aggregated in the admin tab. | `backend/main.py` |
|
|
|
|
| 26 |
| `sessions/` | runtime | Per-session conversation state JSONs. Ephemeral — pruned periodically. Currently includes `anonymous.json` (no-name fallback). |
|
| 27 |
| `policy_facts/` | pre-computed | **256 curated JSONs**, one per policy variant. Each field carries `{value, unit?, source_pdf_path, source_quote}` provenance. The Indian-BFSI-audit-grade machine source; `kb/policies/*.md` are the human-readable mirror. See `_curation_report.md` for the three batches that built it. |
|
| 28 |
| `policies/` | pre-computed | Subfolder per insurer with PDFs / supplementary text used for one-off lookups outside the main ingest pipeline. |
|
| 29 |
+
| `premiums/` | pre-computed | `illustrative_premiums.json` — sample starting premiums pulled from PolicyBazaar / JoinDitto / Beshak + insurer rate cards (2026-05-13). Refreshed by `tools/refresh_premiums.py`. **Illustrative only** per [ADR-007](../70-docs/60-decisions/ADR-007-illustrative-pricing.md). |
|
| 30 |
| `reviews/` | pre-computed | One JSON per insurer with IRDAI claim-settlement metrics, complaints/10K, aggregator sentiment, news tone. Index + leaderboard in `reviews/INDEX.md`. Source: IRDAI Annual Report 2023-24. |
|
| 31 |
|
| 32 |
## Provenance + KPIs
|
|
|
|
| 41 |
|
| 42 |
- [`kb/AUDIT_TRAIL.md`](../kb/AUDIT_TRAIL.md) — end-to-end lineage; `data/policy_facts/` is stage 8 output
|
| 43 |
- [`kb/INDEX.md`](../kb/INDEX.md) — policy index with completeness % per file
|
| 44 |
+
- [ADR-007](../70-docs/60-decisions/ADR-007-illustrative-pricing.md) — pricing is illustrative, never a real quote
|
| 45 |
+
- [ADR-009](../70-docs/60-decisions/ADR-009-19-insurer-comprehensive-schema.md) — 19-insurer scope + 48-field schema
|
40-data/corpus_urls.md
CHANGED
|
@@ -17,18 +17,18 @@
|
|
| 17 |
| star-health | Star Health | Star Cardiac Care Platinum | wordings | https://web.starhealth.in/sites/default/files/policy-clauses/Star-Cardiac-Care-Insurance-Policy_Platinum.pdf | disease-specific |
|
| 18 |
| star-health | Star Health | Star Cancer Care Platinum | wordings | https://web.starhealth.in/sites/default/files/policy-clauses/star-cancer-care-platinum-policy-clauses.pdf | for cancer patients |
|
| 19 |
| star-health | Star Health | Star Hospital Cash | brochure | https://web.starhealth.in/sites/default/files/brochure/hospital_cash_ebrochure_new.pdf | daily-cash benefit |
|
| 20 |
-
| hdfc-ergo | HDFC ERGO | my:Optima Secure | wordings | https://www.hdfcergo.com/docs/default-source/downloads/policy-wordings/health/optima-secure-revision-pw.pdf | flagship indemnity |
|
| 21 |
-
| hdfc-ergo | HDFC ERGO | my:Optima Secure (older variant) | wordings | https://www.hdfcergo.com/docs/default-source/downloads/policy-wordings/health/my-optima-secure_old_pws.pdf | superseded version |
|
| 22 |
| hdfc-ergo | HDFC ERGO | Optima Restore | brochure | https://www.hdfcergo.com/documents/downloads/HEHI/Presales-Set-2/OR-Brochure-Revision-V2.pdf | restoration product |
|
| 23 |
-
| hdfc-ergo | HDFC ERGO | Optima Plus | wordings | https://www.hdfcergo.com/docs/default-source/downloads/policy-wordings/health/optima-plus-policy-wordings.pdf | top-up |
|
| 24 |
-
| hdfc-ergo | HDFC ERGO | Optima Enhance | wordings | https://www.hdfcergo.com/docs/default-source/downloads/policy-wordings/others/optima-enhance-policy-wording.pdf | super top-up |
|
| 25 |
-
| hdfc-ergo | HDFC ERGO | my:health Medisure Prime | wordings | https://www.hdfcergo.com/docs/default-source/downloads/policy-wordings/others/policy-wordings---prime---hrc.pdf | prime indemnity |
|
| 26 |
-
| hdfc-ergo | HDFC ERGO | Energy (Diabetes/Hypertension) | wordings | https://www.hdfcergo.com/docs/default-source/downloads/policy-wordings/health/energy-combined-pw-cis.pdf | combined PW+CIS, disease-specific |
|
| 27 |
-
| hdfc-ergo | HDFC ERGO | Total Health Plan | wordings | https://www.hdfcergo.com/docs/default-source/downloads/policy-wordings/health/total-health-plan--oct-2021.pdf | comprehensive |
|
| 28 |
-
| hdfc-ergo | HDFC ERGO | my:health Suraksha | brochure | https://www.hdfcergo.com/docs/default-source/downloads/prospectus/health/myhealth-suraksha---prospectus.pdf | prospectus (product withdrawn but corpus useful) |
|
| 29 |
-
| hdfc-ergo | HDFC ERGO | my:health Sampoorna Suraksha | brochure | https://www.hdfcergo.com/docs/default-source/downloads/prospectus/health/my_sampoorna_suraksha.pdf | prospectus |
|
| 30 |
-
| hdfc-ergo | HDFC ERGO | my:health Women Suraksha | brochure | https://www.hdfcergo.com/docs/default-source/downloads/brochures/myhealth-women-suraksha-with-premium-table.pdf | women-specific |
|
| 31 |
-
| hdfc-ergo | HDFC ERGO | Group Health Insurance | wordings | https://www.hdfcergo.com/docs/default-source/downloads/policy-wordings/health/group-health-insurance---pw.pdf | group product, structurally similar |
|
| 32 |
| niva-bupa | Niva Bupa | ReAssure 2.0 | wordings | https://transactions.nivabupa.com/pages/doc/policy_wording/ReAssure-2.0-Policy-Wording.pdf | flagship |
|
| 33 |
| niva-bupa | Niva Bupa | ReAssure 3.0 | wordings | https://transactions.nivabupa.com/pages/doc/policy_wording/ReAssure30_Policy_Wordings.pdf | latest gen |
|
| 34 |
| niva-bupa | Niva Bupa | Aspire | wordings | https://transactions.nivabupa.com/pages/doc/policy_wording/Aspire_Policy_Wordings.pdf?v=1.3 | millennials product |
|
|
@@ -48,15 +48,15 @@
|
|
| 48 |
| care-health | Care Health Insurance | Care Advantage | brochure | https://cms.careinsurance.com/cms/public/uploads/download_center/care-advantage-(health-insurance-product)---brochure.pdf | base brochure |
|
| 49 |
| care-health | Care Health Insurance | Care Advantage + add-ons (Protect Plus + Care Shield) | brochure | https://cms.careinsurance.com/cms/public/uploads/download_center/care-advantage-with-add-on-protect-plus-&-care-shield-brochure.pdf | combo brochure |
|
| 50 |
| care-health | Care Health Insurance | Care Heart | brochure | https://cms.careinsurance.com/cms/public/uploads/download_center/care-heart---piano-fold-brochure---web.pdf | cardiac-specific |
|
| 51 |
-
| icici-lombard | ICICI Lombard | Complete Health Insurance (Health Shield) | wordings | https://www.icicilombard.com/docs/default-source/policy-wordings-product-brochure/complete-health-insurance-(health-shield).pdf | retail comprehensive |
|
| 52 |
-
| icici-lombard | ICICI Lombard | Complete Health Insurance (umbrella) | wordings | https://www.icicilombard.com/docs/default-source/default-document-library/icihlip23144v072223-icici-lombard-complete-health-insurance.pdf | umbrella PW |
|
| 53 |
-
| icici-lombard | ICICI Lombard | Health Shield 360 Retail | wordings | https://www.icicilombard.com/docs/default-source/default-document-library/health-shield-360-retail_pw.pdf | 360 retail |
|
| 54 |
-
| icici-lombard | ICICI Lombard | Health Shield 360 Retail | cis | https://www.icicilombard.com/docs/default-source/policy-wordings-product-brochure/health-shield-360-retail---cis.pdf | CIS |
|
| 55 |
-
| icici-lombard | ICICI Lombard | Health Booster (top-up) | wordings | https://www.icicilombard.com/docs/default-source/default-document-library/health-booster_policy-wordings.pdf | top-up |
|
| 56 |
-
| icici-lombard | ICICI Lombard | Health AdvantEdge | wordings | https://www.icicilombard.com/docs/default-source/apps/healthclientapp/assets/pdf/health-advantedge-policy-wordings.pdf | wellness-led product |
|
| 57 |
-
| icici-lombard | ICICI Lombard | Elevate | wordings | https://www.icicilombard.com/docs/default-source/apps/elevateapp/assets/pdf/elevate-policy-wordings.pdf | premium tier |
|
| 58 |
-
| icici-lombard | ICICI Lombard | Health Elite Plus | wordings | https://www.icicilombard.com/docs/default-source/policy-wordings-product-brochure/complete-health-insurance-(health-elite-plus).pdf | elite plus tier |
|
| 59 |
-
| icici-lombard | ICICI Lombard | Arogya Sanjeevani | wordings | https://www.icicilombard.com/docs/default-source/policy-wordings-product-brochure/arogya-sanjeevani-policy-policy-wordings.pdf | IRDAI standard product |
|
| 60 |
| bajaj-allianz | Bajaj Allianz | Health Guard | wordings | https://www.bajajallianz.com/download-documents/health-insurance/health-guard/Health-Guard-Policy-Wordings-print.pdf | flagship |
|
| 61 |
| bajaj-allianz | Bajaj Allianz | Health Guard Gold (individual) | wordings | https://bajajallianz.com/download-documents/health-insurance/health-guard-individual-policy/HG_Gold_Policy_Wording_&_CIS.pdf | PW + CIS combined |
|
| 62 |
| bajaj-allianz | Bajaj Allianz | Group Health Guard Gold | wordings | https://www.bajajallianz.com/download-documents/health-insurance/health-guard-group/Group-HG-Gold-Policy-Wordings.pdf | group |
|
|
@@ -68,14 +68,14 @@
|
|
| 68 |
| bajaj-allianz | Bajaj Allianz | Silver Health | cis | https://www.bajajallianz.com/download-documents/health-insurance/silver-health/silver_health_CIS.pdf | senior citizens (CIS only found) |
|
| 69 |
| bajaj-allianz | Bajaj Allianz | Tax Gain | cis | https://www.bajajallianz.com/download-documents/health-insurance/tax-gain/tax_gain_CIS.pdf | tax-saver plan |
|
| 70 |
| bajaj-allianz | Bajaj Allianz | Group Personal Accident | wordings | https://www.bajajallianz.com/download-documents/health-insurance/GROUP-PERSONAL-ACCIDENT-Policy-Wordings.pdf | PA |
|
| 71 |
-
| new-india | New India Assurance | New India Mediclaim Policy | wordings | https://www.newindia.co.in/assets/docs/know-more/health/new-india-mediclaim-policy/PolicyClauseNewIndiaMediclaimPolicy(NIAHLIP23187V052223).pdf | flagship retail mediclaim |
|
| 72 |
-
| new-india | New India Assurance | New India Mediclaim Policy | brochure | https://www.newindia.co.in/assets/docs/know-more/health/new-india-mediclaim-policy/Prospectus%20New%20India%20Mediclaim%20Policy.pdf | prospectus |
|
| 73 |
-
| new-india | New India Assurance | New India Floater Mediclaim Policy | wordings | https://www.newindia.co.in/assets/docs/know-more/health/floater-mediclaim-policy/Policy%20Clause%20New%20India%20Floater%20Mediclaim%20Policy%20wef%2001%2010%202024.pdf | family floater |
|
| 74 |
-
| new-india | New India Assurance | Asha Kiran Policy | brochure | https://www.newindia.co.in/assets/docs/know-more/health/asha-kiran-policy/Prospectus%20New%20India%20Asha%20Kiran%20Policy%20wef%2001%2004%202021.pdf | women+family hybrid; prospectus |
|
| 75 |
-
| new-india | New India Assurance | Asha Kiran Policy | cis | http://www.newindia.co.in/assets/docs/know-more/health/asha-kiran-policy/Customer%20Information%20Sheet%20NEW%20INDIA%20ASHA%20KIRAN%20POLICY.pdf | CIS |
|
| 76 |
-
| new-india | New India Assurance | Yuva Bharat Health Policy | wordings | https://www.newindia.co.in/assets/docs/know-more/health/yuva-bharat-health-policy/Policy%20Clause%20Yuva%20Bharat%20Health%20Policy%20%20wef%2001%2010%202024_1.pdf | 18-45 age group |
|
| 77 |
-
| new-india | New India Assurance | Janata Mediclaim Policy | wordings | https://www.newindia.co.in/assets/docs/know-more/health/janata-mediclaim-policy/Policy%20Clause%20Janata%20Mediclaim%20Policy.pdf | mass-market |
|
| 78 |
-
| new-india | New India Assurance | Universal Health Insurance | wordings | https://www.newindia.co.in/assets/docs/know-more/health/universal-health-insurance/Policy%20Clause%20Universal%20Health%20Insurance%20Policy.pdf | govt-sponsored |
|
| 79 |
| aditya-birla | Aditya Birla Health Insurance | Activ Health (individual) | wordings | https://www.adityabirlacapital.com/healthinsurance/assets/pdf/policy-wording-form.pdf | flagship; brand may have rebranded as Activ One |
|
| 80 |
| aditya-birla | Aditya Birla Health Insurance | Group Activ Health | wordings | https://www.adityabirlacapital.com/healthinsurance/assets/pdf/new_updated_pdf/Group-Activ-Health-Policy-wordings.pdf | group, structurally similar |
|
| 81 |
| aditya-birla | Aditya Birla Health Insurance | Activ One | brochure | https://www.adityabirlacapital.com/healthinsurance/assets/pdf/active-one/brochure.pdf | newer flagship |
|
|
|
|
| 17 |
| star-health | Star Health | Star Cardiac Care Platinum | wordings | https://web.starhealth.in/sites/default/files/policy-clauses/Star-Cardiac-Care-Insurance-Policy_Platinum.pdf | disease-specific |
|
| 18 |
| star-health | Star Health | Star Cancer Care Platinum | wordings | https://web.starhealth.in/sites/default/files/policy-clauses/star-cancer-care-platinum-policy-clauses.pdf | for cancer patients |
|
| 19 |
| star-health | Star Health | Star Hospital Cash | brochure | https://web.starhealth.in/sites/default/files/brochure/hospital_cash_ebrochure_new.pdf | daily-cash benefit |
|
| 20 |
+
| hdfc-ergo | HDFC ERGO | my:Optima Secure | wordings | https://www.hdfcergo.com/70-docs/default-source/downloads/policy-wordings/health/optima-secure-revision-pw.pdf | flagship indemnity |
|
| 21 |
+
| hdfc-ergo | HDFC ERGO | my:Optima Secure (older variant) | wordings | https://www.hdfcergo.com/70-docs/default-source/downloads/policy-wordings/health/my-optima-secure_old_pws.pdf | superseded version |
|
| 22 |
| hdfc-ergo | HDFC ERGO | Optima Restore | brochure | https://www.hdfcergo.com/documents/downloads/HEHI/Presales-Set-2/OR-Brochure-Revision-V2.pdf | restoration product |
|
| 23 |
+
| hdfc-ergo | HDFC ERGO | Optima Plus | wordings | https://www.hdfcergo.com/70-docs/default-source/downloads/policy-wordings/health/optima-plus-policy-wordings.pdf | top-up |
|
| 24 |
+
| hdfc-ergo | HDFC ERGO | Optima Enhance | wordings | https://www.hdfcergo.com/70-docs/default-source/downloads/policy-wordings/others/optima-enhance-policy-wording.pdf | super top-up |
|
| 25 |
+
| hdfc-ergo | HDFC ERGO | my:health Medisure Prime | wordings | https://www.hdfcergo.com/70-docs/default-source/downloads/policy-wordings/others/policy-wordings---prime---hrc.pdf | prime indemnity |
|
| 26 |
+
| hdfc-ergo | HDFC ERGO | Energy (Diabetes/Hypertension) | wordings | https://www.hdfcergo.com/70-docs/default-source/downloads/policy-wordings/health/energy-combined-pw-cis.pdf | combined PW+CIS, disease-specific |
|
| 27 |
+
| hdfc-ergo | HDFC ERGO | Total Health Plan | wordings | https://www.hdfcergo.com/70-docs/default-source/downloads/policy-wordings/health/total-health-plan--oct-2021.pdf | comprehensive |
|
| 28 |
+
| hdfc-ergo | HDFC ERGO | my:health Suraksha | brochure | https://www.hdfcergo.com/70-docs/default-source/downloads/prospectus/health/myhealth-suraksha---prospectus.pdf | prospectus (product withdrawn but corpus useful) |
|
| 29 |
+
| hdfc-ergo | HDFC ERGO | my:health Sampoorna Suraksha | brochure | https://www.hdfcergo.com/70-docs/default-source/downloads/prospectus/health/my_sampoorna_suraksha.pdf | prospectus |
|
| 30 |
+
| hdfc-ergo | HDFC ERGO | my:health Women Suraksha | brochure | https://www.hdfcergo.com/70-docs/default-source/downloads/brochures/myhealth-women-suraksha-with-premium-table.pdf | women-specific |
|
| 31 |
+
| hdfc-ergo | HDFC ERGO | Group Health Insurance | wordings | https://www.hdfcergo.com/70-docs/default-source/downloads/policy-wordings/health/group-health-insurance---pw.pdf | group product, structurally similar |
|
| 32 |
| niva-bupa | Niva Bupa | ReAssure 2.0 | wordings | https://transactions.nivabupa.com/pages/doc/policy_wording/ReAssure-2.0-Policy-Wording.pdf | flagship |
|
| 33 |
| niva-bupa | Niva Bupa | ReAssure 3.0 | wordings | https://transactions.nivabupa.com/pages/doc/policy_wording/ReAssure30_Policy_Wordings.pdf | latest gen |
|
| 34 |
| niva-bupa | Niva Bupa | Aspire | wordings | https://transactions.nivabupa.com/pages/doc/policy_wording/Aspire_Policy_Wordings.pdf?v=1.3 | millennials product |
|
|
|
|
| 48 |
| care-health | Care Health Insurance | Care Advantage | brochure | https://cms.careinsurance.com/cms/public/uploads/download_center/care-advantage-(health-insurance-product)---brochure.pdf | base brochure |
|
| 49 |
| care-health | Care Health Insurance | Care Advantage + add-ons (Protect Plus + Care Shield) | brochure | https://cms.careinsurance.com/cms/public/uploads/download_center/care-advantage-with-add-on-protect-plus-&-care-shield-brochure.pdf | combo brochure |
|
| 50 |
| care-health | Care Health Insurance | Care Heart | brochure | https://cms.careinsurance.com/cms/public/uploads/download_center/care-heart---piano-fold-brochure---web.pdf | cardiac-specific |
|
| 51 |
+
| icici-lombard | ICICI Lombard | Complete Health Insurance (Health Shield) | wordings | https://www.icicilombard.com/70-docs/default-source/policy-wordings-product-brochure/complete-health-insurance-(health-shield).pdf | retail comprehensive |
|
| 52 |
+
| icici-lombard | ICICI Lombard | Complete Health Insurance (umbrella) | wordings | https://www.icicilombard.com/70-docs/default-source/default-document-library/icihlip23144v072223-icici-lombard-complete-health-insurance.pdf | umbrella PW |
|
| 53 |
+
| icici-lombard | ICICI Lombard | Health Shield 360 Retail | wordings | https://www.icicilombard.com/70-docs/default-source/default-document-library/health-shield-360-retail_pw.pdf | 360 retail |
|
| 54 |
+
| icici-lombard | ICICI Lombard | Health Shield 360 Retail | cis | https://www.icicilombard.com/70-docs/default-source/policy-wordings-product-brochure/health-shield-360-retail---cis.pdf | CIS |
|
| 55 |
+
| icici-lombard | ICICI Lombard | Health Booster (top-up) | wordings | https://www.icicilombard.com/70-docs/default-source/default-document-library/health-booster_policy-wordings.pdf | top-up |
|
| 56 |
+
| icici-lombard | ICICI Lombard | Health AdvantEdge | wordings | https://www.icicilombard.com/70-docs/default-source/apps/healthclientapp/assets/pdf/health-advantedge-policy-wordings.pdf | wellness-led product |
|
| 57 |
+
| icici-lombard | ICICI Lombard | Elevate | wordings | https://www.icicilombard.com/70-docs/default-source/apps/elevateapp/assets/pdf/elevate-policy-wordings.pdf | premium tier |
|
| 58 |
+
| icici-lombard | ICICI Lombard | Health Elite Plus | wordings | https://www.icicilombard.com/70-docs/default-source/policy-wordings-product-brochure/complete-health-insurance-(health-elite-plus).pdf | elite plus tier |
|
| 59 |
+
| icici-lombard | ICICI Lombard | Arogya Sanjeevani | wordings | https://www.icicilombard.com/70-docs/default-source/policy-wordings-product-brochure/arogya-sanjeevani-policy-policy-wordings.pdf | IRDAI standard product |
|
| 60 |
| bajaj-allianz | Bajaj Allianz | Health Guard | wordings | https://www.bajajallianz.com/download-documents/health-insurance/health-guard/Health-Guard-Policy-Wordings-print.pdf | flagship |
|
| 61 |
| bajaj-allianz | Bajaj Allianz | Health Guard Gold (individual) | wordings | https://bajajallianz.com/download-documents/health-insurance/health-guard-individual-policy/HG_Gold_Policy_Wording_&_CIS.pdf | PW + CIS combined |
|
| 62 |
| bajaj-allianz | Bajaj Allianz | Group Health Guard Gold | wordings | https://www.bajajallianz.com/download-documents/health-insurance/health-guard-group/Group-HG-Gold-Policy-Wordings.pdf | group |
|
|
|
|
| 68 |
| bajaj-allianz | Bajaj Allianz | Silver Health | cis | https://www.bajajallianz.com/download-documents/health-insurance/silver-health/silver_health_CIS.pdf | senior citizens (CIS only found) |
|
| 69 |
| bajaj-allianz | Bajaj Allianz | Tax Gain | cis | https://www.bajajallianz.com/download-documents/health-insurance/tax-gain/tax_gain_CIS.pdf | tax-saver plan |
|
| 70 |
| bajaj-allianz | Bajaj Allianz | Group Personal Accident | wordings | https://www.bajajallianz.com/download-documents/health-insurance/GROUP-PERSONAL-ACCIDENT-Policy-Wordings.pdf | PA |
|
| 71 |
+
| new-india | New India Assurance | New India Mediclaim Policy | wordings | https://www.newindia.co.in/assets/70-docs/know-more/health/new-india-mediclaim-policy/PolicyClauseNewIndiaMediclaimPolicy(NIAHLIP23187V052223).pdf | flagship retail mediclaim |
|
| 72 |
+
| new-india | New India Assurance | New India Mediclaim Policy | brochure | https://www.newindia.co.in/assets/70-docs/know-more/health/new-india-mediclaim-policy/Prospectus%20New%20India%20Mediclaim%20Policy.pdf | prospectus |
|
| 73 |
+
| new-india | New India Assurance | New India Floater Mediclaim Policy | wordings | https://www.newindia.co.in/assets/70-docs/know-more/health/floater-mediclaim-policy/Policy%20Clause%20New%20India%20Floater%20Mediclaim%20Policy%20wef%2001%2010%202024.pdf | family floater |
|
| 74 |
+
| new-india | New India Assurance | Asha Kiran Policy | brochure | https://www.newindia.co.in/assets/70-docs/know-more/health/asha-kiran-policy/Prospectus%20New%20India%20Asha%20Kiran%20Policy%20wef%2001%2004%202021.pdf | women+family hybrid; prospectus |
|
| 75 |
+
| new-india | New India Assurance | Asha Kiran Policy | cis | http://www.newindia.co.in/assets/70-docs/know-more/health/asha-kiran-policy/Customer%20Information%20Sheet%20NEW%20INDIA%20ASHA%20KIRAN%20POLICY.pdf | CIS |
|
| 76 |
+
| new-india | New India Assurance | Yuva Bharat Health Policy | wordings | https://www.newindia.co.in/assets/70-docs/know-more/health/yuva-bharat-health-policy/Policy%20Clause%20Yuva%20Bharat%20Health%20Policy%20%20wef%2001%2010%202024_1.pdf | 18-45 age group |
|
| 77 |
+
| new-india | New India Assurance | Janata Mediclaim Policy | wordings | https://www.newindia.co.in/assets/70-docs/know-more/health/janata-mediclaim-policy/Policy%20Clause%20Janata%20Mediclaim%20Policy.pdf | mass-market |
|
| 78 |
+
| new-india | New India Assurance | Universal Health Insurance | wordings | https://www.newindia.co.in/assets/70-docs/know-more/health/universal-health-insurance/Policy%20Clause%20Universal%20Health%20Insurance%20Policy.pdf | govt-sponsored |
|
| 79 |
| aditya-birla | Aditya Birla Health Insurance | Activ Health (individual) | wordings | https://www.adityabirlacapital.com/healthinsurance/assets/pdf/policy-wording-form.pdf | flagship; brand may have rebranded as Activ One |
|
| 80 |
| aditya-birla | Aditya Birla Health Insurance | Group Activ Health | wordings | https://www.adityabirlacapital.com/healthinsurance/assets/pdf/new_updated_pdf/Group-Activ-Health-Policy-wordings.pdf | group, structurally similar |
|
| 81 |
| aditya-birla | Aditya Birla Health Insurance | Activ One | brochure | https://www.adityabirlacapital.com/healthinsurance/assets/pdf/active-one/brochure.pdf | newer flagship |
|
40-data/information_source_map.md
CHANGED
|
@@ -262,7 +262,7 @@ Audited 56 claims — ✅ 0 verified, ⚠️ 47 quote-missing, ❌ 9 broken.
|
|
| 262 |
| `premiums_meta` | `sources_consulted[11]` | ⚠️ url-ok-quote-missing | `https://joinditto.in/articles/health-insurance/niva-bupa-reassure-2-0-premium-chart/` | URL reachable but no source_quote provided |
|
| 263 |
| `premiums_meta` | `sources_consulted[12]` | ⚠️ url-ok-quote-missing | `https://joinditto.in/articles/health-insurance/care-supreme-health-insurance-premium-chart/` | URL reachable but no source_quote provided |
|
| 264 |
| `premiums_meta` | `sources_consulted[13]` | ⚠️ url-ok-quote-missing | `https://joinditto.in/articles/health-insurance/tata-aig-health-insurance-premium-rate-chart/` | URL reachable but no source_quote provided |
|
| 265 |
-
| `premiums_meta` | `sources_consulted[14]` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf` | URL reachable but no source_quote provided |
|
| 266 |
| `premiums_meta` | `sources_consulted[15]` | ⚠️ url-ok-quote-missing | `https://transactions.nivabupa.com/pages/doc/premium_chart/Health%20Premia%20-%20Rate%20Card.pdf` | URL reachable but no source_quote provided |
|
| 267 |
| `premiums_meta` | `sources_consulted[16]` | ⚠️ url-ok-quote-missing | `https://www.scribd.com/document/667873826/Annexure-2-Premium-Chart-Family-Health-Optima-Insurance-Plan-Revised` | URL reachable but no source_quote provided |
|
| 268 |
| `premiums_meta` | `sources_consulted[17]` | ⚠️ url-ok-quote-missing | `https://www.scribd.com/document/682628786/ReAssure-2-0-Rate-Card-with-Tax` | URL reachable but no source_quote provided |
|
|
@@ -292,12 +292,12 @@ Audited 56 claims — ✅ 0 verified, ⚠️ 47 quote-missing, ❌ 9 broken.
|
|
| 292 |
| `niva-bupa__aspire` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.nivabupa.com/family-health-insurance-plans/aspire.html` | URL reachable but no source_quote provided |
|
| 293 |
| `new-india__asha-kiran` | `samples[0].age=30_si=300000` | ⚠️ url-ok-quote-missing | `https://www.newindia.co.in/portal/readMore/Pages/Asha-Kiran-Policy` | URL reachable but no source_quote provided |
|
| 294 |
| `new-india__mediclaim` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.newindia.co.in/portal/readMore/Pages/Mediclaim-Policy` | URL reachable but no source_quote provided |
|
| 295 |
-
| `tata-aig__medicare` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/docs/default-source/default-document-library/medicare-rate-chart-06-12-22.pdf` | URL reachable but no source_quote provided |
|
| 296 |
-
| `tata-aig__medicare` | `samples[1].age=30_si=1000000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/docs/default-source/default-document-library/medicare-rate-chart-06-12-22.pdf` | URL reachable but no source_quote provided |
|
| 297 |
-
| `tata-aig__medicare-premier` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf` | URL reachable but no source_quote provided |
|
| 298 |
-
| `tata-aig__medicare-premier` | `samples[1].age=30_si=1000000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf` | URL reachable but no source_quote provided |
|
| 299 |
-
| `tata-aig__medicare-premier` | `samples[2].age=40_si=500000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf` | URL reachable but no source_quote provided |
|
| 300 |
-
| `tata-aig__medicare-premier` | `samples[3].age=40_si=1000000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf` | URL reachable but no source_quote provided |
|
| 301 |
| `manipalcigna__prohealth-prime-active` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.manipalcigna.com/hospitalization-cover/prohealth-insurance/prohealthprime-active` | URL reachable but no source_quote provided |
|
| 302 |
| `star-health__family-health-optima` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.scribd.com/document/667873826/Annexure-2-Premium-Chart-Family-Health-Optima-Insurance-Plan-Revised` | URL reachable but no source_quote provided |
|
| 303 |
| `star-health__family-health-optima` | `samples[1].age=40_si=500000` | ⚠️ url-ok-quote-missing | `https://www.scribd.com/document/667873826/Annexure-2-Premium-Chart-Family-Health-Optima-Insurance-Plan-Revised` | URL reachable but no source_quote provided |
|
|
|
|
| 262 |
| `premiums_meta` | `sources_consulted[11]` | ⚠️ url-ok-quote-missing | `https://joinditto.in/articles/health-insurance/niva-bupa-reassure-2-0-premium-chart/` | URL reachable but no source_quote provided |
|
| 263 |
| `premiums_meta` | `sources_consulted[12]` | ⚠️ url-ok-quote-missing | `https://joinditto.in/articles/health-insurance/care-supreme-health-insurance-premium-chart/` | URL reachable but no source_quote provided |
|
| 264 |
| `premiums_meta` | `sources_consulted[13]` | ⚠️ url-ok-quote-missing | `https://joinditto.in/articles/health-insurance/tata-aig-health-insurance-premium-rate-chart/` | URL reachable but no source_quote provided |
|
| 265 |
+
| `premiums_meta` | `sources_consulted[14]` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/70-docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf` | URL reachable but no source_quote provided |
|
| 266 |
| `premiums_meta` | `sources_consulted[15]` | ⚠️ url-ok-quote-missing | `https://transactions.nivabupa.com/pages/doc/premium_chart/Health%20Premia%20-%20Rate%20Card.pdf` | URL reachable but no source_quote provided |
|
| 267 |
| `premiums_meta` | `sources_consulted[16]` | ⚠️ url-ok-quote-missing | `https://www.scribd.com/document/667873826/Annexure-2-Premium-Chart-Family-Health-Optima-Insurance-Plan-Revised` | URL reachable but no source_quote provided |
|
| 268 |
| `premiums_meta` | `sources_consulted[17]` | ⚠️ url-ok-quote-missing | `https://www.scribd.com/document/682628786/ReAssure-2-0-Rate-Card-with-Tax` | URL reachable but no source_quote provided |
|
|
|
|
| 292 |
| `niva-bupa__aspire` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.nivabupa.com/family-health-insurance-plans/aspire.html` | URL reachable but no source_quote provided |
|
| 293 |
| `new-india__asha-kiran` | `samples[0].age=30_si=300000` | ⚠️ url-ok-quote-missing | `https://www.newindia.co.in/portal/readMore/Pages/Asha-Kiran-Policy` | URL reachable but no source_quote provided |
|
| 294 |
| `new-india__mediclaim` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.newindia.co.in/portal/readMore/Pages/Mediclaim-Policy` | URL reachable but no source_quote provided |
|
| 295 |
+
| `tata-aig__medicare` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/70-docs/default-source/default-document-library/medicare-rate-chart-06-12-22.pdf` | URL reachable but no source_quote provided |
|
| 296 |
+
| `tata-aig__medicare` | `samples[1].age=30_si=1000000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/70-docs/default-source/default-document-library/medicare-rate-chart-06-12-22.pdf` | URL reachable but no source_quote provided |
|
| 297 |
+
| `tata-aig__medicare-premier` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/70-docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf` | URL reachable but no source_quote provided |
|
| 298 |
+
| `tata-aig__medicare-premier` | `samples[1].age=30_si=1000000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/70-docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf` | URL reachable but no source_quote provided |
|
| 299 |
+
| `tata-aig__medicare-premier` | `samples[2].age=40_si=500000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/70-docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf` | URL reachable but no source_quote provided |
|
| 300 |
+
| `tata-aig__medicare-premier` | `samples[3].age=40_si=1000000` | ⚠️ url-ok-quote-missing | `https://www.axisbank.com/70-docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf` | URL reachable but no source_quote provided |
|
| 301 |
| `manipalcigna__prohealth-prime-active` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.manipalcigna.com/hospitalization-cover/prohealth-insurance/prohealthprime-active` | URL reachable but no source_quote provided |
|
| 302 |
| `star-health__family-health-optima` | `samples[0].age=30_si=500000` | ⚠️ url-ok-quote-missing | `https://www.scribd.com/document/667873826/Annexure-2-Premium-Chart-Family-Health-Optima-Insurance-Plan-Revised` | URL reachable but no source_quote provided |
|
| 303 |
| `star-health__family-health-optima` | `samples[1].age=40_si=500000` | ⚠️ url-ok-quote-missing | `https://www.scribd.com/document/667873826/Annexure-2-Premium-Chart-Family-Health-Optima-Insurance-Plan-Revised` | URL reachable but no source_quote provided |
|
40-data/llm_health.json
CHANGED
|
@@ -1,125 +1,145 @@
|
|
| 1 |
{
|
| 2 |
-
"updated_at": "2026-05-
|
| 3 |
"models": {
|
| 4 |
"qwen/qwen3-next-80b-a3b-instruct": {
|
| 5 |
"model": "qwen/qwen3-next-80b-a3b-instruct",
|
| 6 |
"status": "healthy",
|
| 7 |
-
"last_success_at": "2026-05-
|
| 8 |
"last_failure_at": null,
|
| 9 |
"last_error": null,
|
| 10 |
-
"latency_ms":
|
| 11 |
"consecutive_failures": 0,
|
| 12 |
-
"tested_at": "2026-05-
|
| 13 |
},
|
| 14 |
"qwen/qwen3.5-122b-a10b": {
|
| 15 |
"model": "qwen/qwen3.5-122b-a10b",
|
| 16 |
-
"status": "
|
| 17 |
"last_success_at": null,
|
| 18 |
-
"last_failure_at": "2026-05-
|
| 19 |
-
"last_error": "
|
| 20 |
"latency_ms": null,
|
| 21 |
-
"consecutive_failures":
|
| 22 |
-
"tested_at": "2026-05-
|
| 23 |
},
|
| 24 |
"openai/gpt-oss-120b": {
|
| 25 |
"model": "openai/gpt-oss-120b",
|
| 26 |
-
"status": "
|
| 27 |
"last_success_at": null,
|
| 28 |
-
"last_failure_at": "2026-05-
|
| 29 |
"last_error": "empty_content",
|
| 30 |
"latency_ms": null,
|
| 31 |
-
"consecutive_failures":
|
| 32 |
-
"tested_at": "2026-05-
|
| 33 |
},
|
| 34 |
"mistralai/mistral-large-3-675b-instruct-2512": {
|
| 35 |
"model": "mistralai/mistral-large-3-675b-instruct-2512",
|
| 36 |
-
"status": "
|
| 37 |
"last_success_at": null,
|
| 38 |
-
"last_failure_at": "2026-05-
|
| 39 |
"last_error": "timeout",
|
| 40 |
"latency_ms": null,
|
| 41 |
-
"consecutive_failures":
|
| 42 |
-
"tested_at": "2026-05-
|
| 43 |
},
|
| 44 |
"nvidia/llama-3.3-nemotron-super-49b-v1.5": {
|
| 45 |
"model": "nvidia/llama-3.3-nemotron-super-49b-v1.5",
|
| 46 |
-
"status": "
|
| 47 |
"last_success_at": null,
|
| 48 |
-
"last_failure_at": "2026-05-
|
| 49 |
"last_error": "empty_content",
|
| 50 |
"latency_ms": null,
|
| 51 |
-
"consecutive_failures":
|
| 52 |
-
"tested_at": "2026-05-
|
| 53 |
},
|
| 54 |
"meta/llama-3.3-70b-instruct": {
|
| 55 |
"model": "meta/llama-3.3-70b-instruct",
|
| 56 |
-
"status": "
|
| 57 |
-
"last_success_at": "2026-05-
|
| 58 |
-
"last_failure_at":
|
| 59 |
"last_error": null,
|
| 60 |
-
"latency_ms":
|
| 61 |
"consecutive_failures": 0,
|
| 62 |
-
"tested_at": "2026-05-
|
| 63 |
},
|
| 64 |
"deepseek-ai/deepseek-v4-pro": {
|
| 65 |
"model": "deepseek-ai/deepseek-v4-pro",
|
| 66 |
-
"status": "
|
| 67 |
"last_success_at": null,
|
| 68 |
-
"last_failure_at": "2026-05-
|
| 69 |
"last_error": "timeout",
|
| 70 |
"latency_ms": null,
|
| 71 |
-
"consecutive_failures":
|
| 72 |
-
"tested_at": "2026-05-
|
| 73 |
},
|
| 74 |
"nvidia/nemotron-3-nano-30b-a3b": {
|
| 75 |
"model": "nvidia/nemotron-3-nano-30b-a3b",
|
| 76 |
-
"status": "
|
| 77 |
"last_success_at": null,
|
| 78 |
-
"last_failure_at": "2026-05-
|
| 79 |
"last_error": "empty_content",
|
| 80 |
"latency_ms": null,
|
| 81 |
-
"consecutive_failures":
|
| 82 |
-
"tested_at": "2026-05-
|
| 83 |
},
|
| 84 |
"deepseek-ai/deepseek-v4-flash": {
|
| 85 |
"model": "deepseek-ai/deepseek-v4-flash",
|
| 86 |
-
"status": "
|
| 87 |
"last_success_at": null,
|
| 88 |
-
"last_failure_at": "2026-05-
|
| 89 |
"last_error": "timeout",
|
| 90 |
"latency_ms": null,
|
| 91 |
-
"consecutive_failures":
|
| 92 |
-
"tested_at": "2026-05-
|
| 93 |
},
|
| 94 |
"moonshotai/kimi-k2-instruct-0905": {
|
| 95 |
"model": "moonshotai/kimi-k2-instruct-0905",
|
| 96 |
-
"status": "
|
| 97 |
"last_success_at": null,
|
| 98 |
-
"last_failure_at": "2026-05-
|
| 99 |
"last_error": "http_404",
|
| 100 |
"latency_ms": null,
|
| 101 |
-
"consecutive_failures":
|
| 102 |
-
"tested_at": "2026-05-
|
| 103 |
},
|
| 104 |
"minimaxai/minimax-m2.5": {
|
| 105 |
"model": "minimaxai/minimax-m2.5",
|
| 106 |
-
"status": "
|
| 107 |
"last_success_at": null,
|
| 108 |
-
"last_failure_at": "2026-05-
|
| 109 |
"last_error": "http_410",
|
| 110 |
"latency_ms": null,
|
| 111 |
-
"consecutive_failures":
|
| 112 |
-
"tested_at": "2026-05-
|
| 113 |
},
|
| 114 |
"meta/llama-4-maverick-17b-128e-instruct": {
|
| 115 |
"model": "meta/llama-4-maverick-17b-128e-instruct",
|
| 116 |
"status": "degraded",
|
| 117 |
-
"last_success_at":
|
| 118 |
"last_failure_at": "2026-05-14T01:03:07Z",
|
| 119 |
-
"last_error":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 120 |
"latency_ms": null,
|
| 121 |
-
"consecutive_failures":
|
| 122 |
-
"tested_at": "2026-05-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 123 |
}
|
| 124 |
}
|
| 125 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"updated_at": "2026-05-14T16:35:39Z",
|
| 3 |
"models": {
|
| 4 |
"qwen/qwen3-next-80b-a3b-instruct": {
|
| 5 |
"model": "qwen/qwen3-next-80b-a3b-instruct",
|
| 6 |
"status": "healthy",
|
| 7 |
+
"last_success_at": "2026-05-14T16:35:39Z",
|
| 8 |
"last_failure_at": null,
|
| 9 |
"last_error": null,
|
| 10 |
+
"latency_ms": 286,
|
| 11 |
"consecutive_failures": 0,
|
| 12 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 13 |
},
|
| 14 |
"qwen/qwen3.5-122b-a10b": {
|
| 15 |
"model": "qwen/qwen3.5-122b-a10b",
|
| 16 |
+
"status": "down",
|
| 17 |
"last_success_at": null,
|
| 18 |
+
"last_failure_at": "2026-05-14T16:35:39Z",
|
| 19 |
+
"last_error": "timeout",
|
| 20 |
"latency_ms": null,
|
| 21 |
+
"consecutive_failures": 5,
|
| 22 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 23 |
},
|
| 24 |
"openai/gpt-oss-120b": {
|
| 25 |
"model": "openai/gpt-oss-120b",
|
| 26 |
+
"status": "down",
|
| 27 |
"last_success_at": null,
|
| 28 |
+
"last_failure_at": "2026-05-14T16:35:39Z",
|
| 29 |
"last_error": "empty_content",
|
| 30 |
"latency_ms": null,
|
| 31 |
+
"consecutive_failures": 5,
|
| 32 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 33 |
},
|
| 34 |
"mistralai/mistral-large-3-675b-instruct-2512": {
|
| 35 |
"model": "mistralai/mistral-large-3-675b-instruct-2512",
|
| 36 |
+
"status": "down",
|
| 37 |
"last_success_at": null,
|
| 38 |
+
"last_failure_at": "2026-05-14T16:35:39Z",
|
| 39 |
"last_error": "timeout",
|
| 40 |
"latency_ms": null,
|
| 41 |
+
"consecutive_failures": 5,
|
| 42 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 43 |
},
|
| 44 |
"nvidia/llama-3.3-nemotron-super-49b-v1.5": {
|
| 45 |
"model": "nvidia/llama-3.3-nemotron-super-49b-v1.5",
|
| 46 |
+
"status": "down",
|
| 47 |
"last_success_at": null,
|
| 48 |
+
"last_failure_at": "2026-05-14T16:35:39Z",
|
| 49 |
"last_error": "empty_content",
|
| 50 |
"latency_ms": null,
|
| 51 |
+
"consecutive_failures": 5,
|
| 52 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 53 |
},
|
| 54 |
"meta/llama-3.3-70b-instruct": {
|
| 55 |
"model": "meta/llama-3.3-70b-instruct",
|
| 56 |
+
"status": "healthy",
|
| 57 |
+
"last_success_at": "2026-05-14T16:35:39Z",
|
| 58 |
+
"last_failure_at": "2026-05-14T15:58:40Z",
|
| 59 |
"last_error": null,
|
| 60 |
+
"latency_ms": 479,
|
| 61 |
"consecutive_failures": 0,
|
| 62 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 63 |
},
|
| 64 |
"deepseek-ai/deepseek-v4-pro": {
|
| 65 |
"model": "deepseek-ai/deepseek-v4-pro",
|
| 66 |
+
"status": "down",
|
| 67 |
"last_success_at": null,
|
| 68 |
+
"last_failure_at": "2026-05-14T16:35:39Z",
|
| 69 |
"last_error": "timeout",
|
| 70 |
"latency_ms": null,
|
| 71 |
+
"consecutive_failures": 5,
|
| 72 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 73 |
},
|
| 74 |
"nvidia/nemotron-3-nano-30b-a3b": {
|
| 75 |
"model": "nvidia/nemotron-3-nano-30b-a3b",
|
| 76 |
+
"status": "down",
|
| 77 |
"last_success_at": null,
|
| 78 |
+
"last_failure_at": "2026-05-14T16:35:39Z",
|
| 79 |
"last_error": "empty_content",
|
| 80 |
"latency_ms": null,
|
| 81 |
+
"consecutive_failures": 5,
|
| 82 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 83 |
},
|
| 84 |
"deepseek-ai/deepseek-v4-flash": {
|
| 85 |
"model": "deepseek-ai/deepseek-v4-flash",
|
| 86 |
+
"status": "down",
|
| 87 |
"last_success_at": null,
|
| 88 |
+
"last_failure_at": "2026-05-14T16:35:39Z",
|
| 89 |
"last_error": "timeout",
|
| 90 |
"latency_ms": null,
|
| 91 |
+
"consecutive_failures": 5,
|
| 92 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 93 |
},
|
| 94 |
"moonshotai/kimi-k2-instruct-0905": {
|
| 95 |
"model": "moonshotai/kimi-k2-instruct-0905",
|
| 96 |
+
"status": "down",
|
| 97 |
"last_success_at": null,
|
| 98 |
+
"last_failure_at": "2026-05-14T16:35:39Z",
|
| 99 |
"last_error": "http_404",
|
| 100 |
"latency_ms": null,
|
| 101 |
+
"consecutive_failures": 5,
|
| 102 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 103 |
},
|
| 104 |
"minimaxai/minimax-m2.5": {
|
| 105 |
"model": "minimaxai/minimax-m2.5",
|
| 106 |
+
"status": "down",
|
| 107 |
"last_success_at": null,
|
| 108 |
+
"last_failure_at": "2026-05-14T16:35:39Z",
|
| 109 |
"last_error": "http_410",
|
| 110 |
"latency_ms": null,
|
| 111 |
+
"consecutive_failures": 5,
|
| 112 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 113 |
},
|
| 114 |
"meta/llama-4-maverick-17b-128e-instruct": {
|
| 115 |
"model": "meta/llama-4-maverick-17b-128e-instruct",
|
| 116 |
"status": "degraded",
|
| 117 |
+
"last_success_at": "2026-05-14T16:35:39Z",
|
| 118 |
"last_failure_at": "2026-05-14T01:03:07Z",
|
| 119 |
+
"last_error": null,
|
| 120 |
+
"latency_ms": 7197,
|
| 121 |
+
"consecutive_failures": 0,
|
| 122 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 123 |
+
},
|
| 124 |
+
"openrouter:openai/gpt-oss-120b": {
|
| 125 |
+
"model": "openrouter:openai/gpt-oss-120b",
|
| 126 |
+
"status": "down",
|
| 127 |
+
"last_success_at": null,
|
| 128 |
+
"last_failure_at": "2026-05-14T16:35:39Z",
|
| 129 |
+
"last_error": "empty_content",
|
| 130 |
"latency_ms": null,
|
| 131 |
+
"consecutive_failures": 4,
|
| 132 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 133 |
+
},
|
| 134 |
+
"groq:llama-3.3-70b-versatile": {
|
| 135 |
+
"model": "groq:llama-3.3-70b-versatile",
|
| 136 |
+
"status": "healthy",
|
| 137 |
+
"last_success_at": "2026-05-14T16:35:39Z",
|
| 138 |
+
"last_failure_at": null,
|
| 139 |
+
"last_error": null,
|
| 140 |
+
"latency_ms": 200,
|
| 141 |
+
"consecutive_failures": 0,
|
| 142 |
+
"tested_at": "2026-05-14T16:35:39Z"
|
| 143 |
}
|
| 144 |
}
|
| 145 |
}
|
40-data/llm_usage.jsonl
ADDED
|
@@ -0,0 +1,313 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"ts": "2026-05-14T10:15:50Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31360, "success": true}
|
| 2 |
+
{"ts": "2026-05-14T10:15:59Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 9040, "success": true}
|
| 3 |
+
{"ts": "2026-05-14T10:16:13Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 13156, "success": true}
|
| 4 |
+
{"ts": "2026-05-14T10:16:47Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31769, "success": true}
|
| 5 |
+
{"ts": "2026-05-14T10:17:15Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 15073, "success": true}
|
| 6 |
+
{"ts": "2026-05-14T10:17:31Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 16104, "success": true}
|
| 7 |
+
{"ts": "2026-05-14T10:18:03Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31495, "success": true}
|
| 8 |
+
{"ts": "2026-05-14T10:18:22Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 7038, "success": true}
|
| 9 |
+
{"ts": "2026-05-14T10:18:32Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 7555, "success": true}
|
| 10 |
+
{"ts": "2026-05-14T10:18:40Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 8097, "success": true}
|
| 11 |
+
{"ts": "2026-05-14T10:18:50Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 8971, "success": true}
|
| 12 |
+
{"ts": "2026-05-14T10:19:02Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1537, "success": true}
|
| 13 |
+
{"ts": "2026-05-14T10:19:10Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 7407, "success": true}
|
| 14 |
+
{"ts": "2026-05-14T10:19:24Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2668, "success": true}
|
| 15 |
+
{"ts": "2026-05-14T10:19:55Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31366, "success": true}
|
| 16 |
+
{"ts": "2026-05-14T10:20:05Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4099, "success": true}
|
| 17 |
+
{"ts": "2026-05-14T10:20:23Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 18564, "success": true}
|
| 18 |
+
{"ts": "2026-05-14T10:20:30Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1260, "success": true}
|
| 19 |
+
{"ts": "2026-05-14T10:20:41Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 11219, "success": true}
|
| 20 |
+
{"ts": "2026-05-14T10:20:49Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3964, "success": true}
|
| 21 |
+
{"ts": "2026-05-14T10:20:56Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 7004, "success": true}
|
| 22 |
+
{"ts": "2026-05-14T10:21:05Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4454, "success": true}
|
| 23 |
+
{"ts": "2026-05-14T10:21:13Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1802, "success": true}
|
| 24 |
+
{"ts": "2026-05-14T10:21:20Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2258, "success": true}
|
| 25 |
+
{"ts": "2026-05-14T10:21:26Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2145, "success": true}
|
| 26 |
+
{"ts": "2026-05-14T10:21:31Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 4520, "success": true}
|
| 27 |
+
{"ts": "2026-05-14T10:21:46Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4123, "success": true}
|
| 28 |
+
{"ts": "2026-05-14T10:22:09Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 23749, "success": true}
|
| 29 |
+
{"ts": "2026-05-14T10:22:17Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2506, "success": true}
|
| 30 |
+
{"ts": "2026-05-14T10:22:27Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 10569, "success": true}
|
| 31 |
+
{"ts": "2026-05-14T10:22:36Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1997, "success": true}
|
| 32 |
+
{"ts": "2026-05-14T10:22:49Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1465, "success": true}
|
| 33 |
+
{"ts": "2026-05-14T10:22:56Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2554, "success": true}
|
| 34 |
+
{"ts": "2026-05-14T10:23:14Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5818, "success": true}
|
| 35 |
+
{"ts": "2026-05-14T10:23:21Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 6689, "success": true}
|
| 36 |
+
{"ts": "2026-05-14T10:23:33Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 7966, "success": true}
|
| 37 |
+
{"ts": "2026-05-14T10:23:39Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 5209, "success": true}
|
| 38 |
+
{"ts": "2026-05-14T10:23:50Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3082, "success": true}
|
| 39 |
+
{"ts": "2026-05-14T10:24:00Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 9626, "success": true}
|
| 40 |
+
{"ts": "2026-05-14T10:24:06Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1849, "success": true}
|
| 41 |
+
{"ts": "2026-05-14T10:24:39Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 32852, "success": true}
|
| 42 |
+
{"ts": "2026-05-14T10:24:58Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5131, "success": true}
|
| 43 |
+
{"ts": "2026-05-14T10:25:29Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31738, "success": true}
|
| 44 |
+
{"ts": "2026-05-14T10:25:39Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4841, "success": true}
|
| 45 |
+
{"ts": "2026-05-14T10:25:54Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 15091, "success": true}
|
| 46 |
+
{"ts": "2026-05-14T10:26:04Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4970, "success": true}
|
| 47 |
+
{"ts": "2026-05-14T10:26:15Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 11212, "success": true}
|
| 48 |
+
{"ts": "2026-05-14T10:26:22Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2451, "success": true}
|
| 49 |
+
{"ts": "2026-05-14T10:26:37Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 14432, "success": true}
|
| 50 |
+
{"ts": "2026-05-14T10:26:46Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3246, "success": true}
|
| 51 |
+
{"ts": "2026-05-14T10:26:59Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 12709, "success": true}
|
| 52 |
+
{"ts": "2026-05-14T10:27:05Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1996, "success": true}
|
| 53 |
+
{"ts": "2026-05-14T10:27:16Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3101, "success": true}
|
| 54 |
+
{"ts": "2026-05-14T10:29:32Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2877, "success": true}
|
| 55 |
+
{"ts": "2026-05-14T10:29:45Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 7957, "success": true}
|
| 56 |
+
{"ts": "2026-05-14T10:30:17Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31909, "success": true}
|
| 57 |
+
{"ts": "2026-05-14T10:30:29Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5521, "success": true}
|
| 58 |
+
{"ts": "2026-05-14T10:31:01Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 32362, "success": true}
|
| 59 |
+
{"ts": "2026-05-14T10:31:10Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4203, "success": true}
|
| 60 |
+
{"ts": "2026-05-14T10:31:20Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 9069, "success": true}
|
| 61 |
+
{"ts": "2026-05-14T10:31:35Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 9179, "success": true}
|
| 62 |
+
{"ts": "2026-05-14T10:31:47Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 11771, "success": true}
|
| 63 |
+
{"ts": "2026-05-14T10:31:58Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1771, "success": true}
|
| 64 |
+
{"ts": "2026-05-14T10:32:09Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 10849, "success": true}
|
| 65 |
+
{"ts": "2026-05-14T10:32:20Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 6659, "success": true}
|
| 66 |
+
{"ts": "2026-05-14T10:32:30Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 10241, "success": true}
|
| 67 |
+
{"ts": "2026-05-14T10:32:42Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4992, "success": true}
|
| 68 |
+
{"ts": "2026-05-14T10:33:14Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31934, "success": true}
|
| 69 |
+
{"ts": "2026-05-14T10:33:33Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5597, "success": true}
|
| 70 |
+
{"ts": "2026-05-14T10:33:44Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 10698, "success": true}
|
| 71 |
+
{"ts": "2026-05-14T10:34:02Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 6989, "success": true}
|
| 72 |
+
{"ts": "2026-05-14T10:34:26Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2838, "success": true}
|
| 73 |
+
{"ts": "2026-05-14T10:34:36Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3816, "success": true}
|
| 74 |
+
{"ts": "2026-05-14T10:34:58Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 9211, "success": true}
|
| 75 |
+
{"ts": "2026-05-14T10:35:06Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2517, "success": true}
|
| 76 |
+
{"ts": "2026-05-14T10:35:16Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 10844, "success": true}
|
| 77 |
+
{"ts": "2026-05-14T10:35:32Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2960, "success": true}
|
| 78 |
+
{"ts": "2026-05-14T10:35:42Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 10192, "success": true}
|
| 79 |
+
{"ts": "2026-05-14T10:35:49Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2934, "success": true}
|
| 80 |
+
{"ts": "2026-05-14T10:36:05Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 15610, "success": true}
|
| 81 |
+
{"ts": "2026-05-14T10:36:36Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 11783, "success": true}
|
| 82 |
+
{"ts": "2026-05-14T10:36:43Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 7276, "success": true}
|
| 83 |
+
{"ts": "2026-05-14T10:36:52Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4211, "success": true}
|
| 84 |
+
{"ts": "2026-05-14T10:37:06Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 13798, "success": true}
|
| 85 |
+
{"ts": "2026-05-14T10:37:14Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2844, "success": true}
|
| 86 |
+
{"ts": "2026-05-14T10:37:28Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 13921, "success": true}
|
| 87 |
+
{"ts": "2026-05-14T10:37:35Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2468, "success": true}
|
| 88 |
+
{"ts": "2026-05-14T10:37:43Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 7914, "success": true}
|
| 89 |
+
{"ts": "2026-05-14T10:37:53Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5109, "success": true}
|
| 90 |
+
{"ts": "2026-05-14T10:38:00Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 6859, "success": true}
|
| 91 |
+
{"ts": "2026-05-14T10:38:15Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2874, "success": true}
|
| 92 |
+
{"ts": "2026-05-14T10:38:22Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3086, "success": true}
|
| 93 |
+
{"ts": "2026-05-14T10:38:41Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 13388, "success": true}
|
| 94 |
+
{"ts": "2026-05-14T10:38:49Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1701, "success": true}
|
| 95 |
+
{"ts": "2026-05-14T10:38:55Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 6752, "success": true}
|
| 96 |
+
{"ts": "2026-05-14T10:39:11Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4275, "success": true}
|
| 97 |
+
{"ts": "2026-05-14T10:39:43Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31996, "success": true}
|
| 98 |
+
{"ts": "2026-05-14T10:39:57Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2791, "success": true}
|
| 99 |
+
{"ts": "2026-05-14T10:40:10Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 7170, "success": true}
|
| 100 |
+
{"ts": "2026-05-14T10:40:30Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3013, "success": true}
|
| 101 |
+
{"ts": "2026-05-14T10:40:37Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 7066, "success": true}
|
| 102 |
+
{"ts": "2026-05-14T10:40:49Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4538, "success": true}
|
| 103 |
+
{"ts": "2026-05-14T10:41:02Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 12738, "success": true}
|
| 104 |
+
{"ts": "2026-05-14T10:41:19Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 10598, "success": true}
|
| 105 |
+
{"ts": "2026-05-14T10:41:37Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 17983, "success": true}
|
| 106 |
+
{"ts": "2026-05-14T10:41:45Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2419, "success": true}
|
| 107 |
+
{"ts": "2026-05-14T10:42:23Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 38599, "success": true}
|
| 108 |
+
{"ts": "2026-05-14T10:42:44Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 8366, "success": true}
|
| 109 |
+
{"ts": "2026-05-14T10:42:53Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3690, "success": true}
|
| 110 |
+
{"ts": "2026-05-14T10:42:59Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1886, "success": true}
|
| 111 |
+
{"ts": "2026-05-14T10:43:10Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4478, "success": true}
|
| 112 |
+
{"ts": "2026-05-14T10:43:41Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31793, "success": true}
|
| 113 |
+
{"ts": "2026-05-14T10:43:59Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4351, "success": true}
|
| 114 |
+
{"ts": "2026-05-14T10:44:30Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31824, "success": true}
|
| 115 |
+
{"ts": "2026-05-14T10:44:42Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 6407, "success": true}
|
| 116 |
+
{"ts": "2026-05-14T10:45:14Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31842, "success": true}
|
| 117 |
+
{"ts": "2026-05-14T10:45:21Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2263, "success": true}
|
| 118 |
+
{"ts": "2026-05-14T10:45:55Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 33995, "success": true}
|
| 119 |
+
{"ts": "2026-05-14T10:46:02Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2543, "success": true}
|
| 120 |
+
{"ts": "2026-05-14T10:46:14Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "mistralai/mistral-large-3-675b-instruct-2512", "latency_ms": 11744, "success": true}
|
| 121 |
+
{"ts": "2026-05-14T10:46:25Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3612, "success": true}
|
| 122 |
+
{"ts": "2026-05-14T10:46:57Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31815, "success": true}
|
| 123 |
+
{"ts": "2026-05-14T10:47:07Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3819, "success": true}
|
| 124 |
+
{"ts": "2026-05-14T10:47:39Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 32108, "success": true}
|
| 125 |
+
{"ts": "2026-05-14T10:47:46Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2600, "success": true}
|
| 126 |
+
{"ts": "2026-05-14T10:48:18Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31845, "success": true}
|
| 127 |
+
{"ts": "2026-05-14T10:48:28Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3472, "success": true}
|
| 128 |
+
{"ts": "2026-05-14T10:48:36Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3733, "success": true}
|
| 129 |
+
{"ts": "2026-05-14T10:48:51Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 9824, "success": true}
|
| 130 |
+
{"ts": "2026-05-14T10:49:01Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4267, "success": true}
|
| 131 |
+
{"ts": "2026-05-14T10:49:33Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31637, "success": true}
|
| 132 |
+
{"ts": "2026-05-14T10:49:44Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2963, "success": true}
|
| 133 |
+
{"ts": "2026-05-14T10:50:16Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31683, "success": true}
|
| 134 |
+
{"ts": "2026-05-14T10:50:38Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 10257, "success": true}
|
| 135 |
+
{"ts": "2026-05-14T10:51:10Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31804, "success": true}
|
| 136 |
+
{"ts": "2026-05-14T10:51:19Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4728, "success": true}
|
| 137 |
+
{"ts": "2026-05-14T10:51:51Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31592, "success": true}
|
| 138 |
+
{"ts": "2026-05-14T10:52:05Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3106, "success": true}
|
| 139 |
+
{"ts": "2026-05-14T10:52:37Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31786, "success": true}
|
| 140 |
+
{"ts": "2026-05-14T10:52:49Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3904, "success": true}
|
| 141 |
+
{"ts": "2026-05-14T10:53:20Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31818, "success": true}
|
| 142 |
+
{"ts": "2026-05-14T10:53:29Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4181, "success": true}
|
| 143 |
+
{"ts": "2026-05-14T10:53:40Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2611, "success": true}
|
| 144 |
+
{"ts": "2026-05-14T10:53:52Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2774, "success": true}
|
| 145 |
+
{"ts": "2026-05-14T10:56:10Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 9953, "success": true}
|
| 146 |
+
{"ts": "2026-05-14T10:56:18Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3271, "success": true}
|
| 147 |
+
{"ts": "2026-05-14T10:56:26Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4076, "success": true}
|
| 148 |
+
{"ts": "2026-05-14T10:56:39Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5115, "success": true}
|
| 149 |
+
{"ts": "2026-05-14T10:56:52Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 7039, "success": true}
|
| 150 |
+
{"ts": "2026-05-14T15:56:21Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3867, "success": true}
|
| 151 |
+
{"ts": "2026-05-14T15:56:53Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31588, "success": true}
|
| 152 |
+
{"ts": "2026-05-14T15:57:03Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 6039, "success": true}
|
| 153 |
+
{"ts": "2026-05-14T15:58:30Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5397, "success": true}
|
| 154 |
+
{"ts": "2026-05-14T15:58:30Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 14418, "success": true}
|
| 155 |
+
{"ts": "2026-05-14T15:58:35Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 27521, "success": true}
|
| 156 |
+
{"ts": "2026-05-14T15:58:35Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 14672, "success": true}
|
| 157 |
+
{"ts": "2026-05-14T15:58:35Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 23267, "success": true}
|
| 158 |
+
{"ts": "2026-05-14T15:58:40Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4862, "success": true}
|
| 159 |
+
{"ts": "2026-05-14T15:58:40Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": null, "latency_ms": 36706, "success": false}
|
| 160 |
+
{"ts": "2026-05-14T15:58:50Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 9963, "success": true}
|
| 161 |
+
{"ts": "2026-05-14T15:58:56Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 12080, "success": true}
|
| 162 |
+
{"ts": "2026-05-14T15:58:57Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 7960, "success": true}
|
| 163 |
+
{"ts": "2026-05-14T15:59:07Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31821, "success": true}
|
| 164 |
+
{"ts": "2026-05-14T15:59:07Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 32674, "success": true}
|
| 165 |
+
{"ts": "2026-05-14T15:59:29Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 39114, "success": true}
|
| 166 |
+
{"ts": "2026-05-14T15:59:33Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 36636, "success": true}
|
| 167 |
+
{"ts": "2026-05-14T15:59:33Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 6379, "success": true}
|
| 168 |
+
{"ts": "2026-05-14T15:59:33Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 36876, "success": true}
|
| 169 |
+
{"ts": "2026-05-14T15:59:33Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 10677, "success": true}
|
| 170 |
+
{"ts": "2026-05-14T15:59:38Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5173, "success": true}
|
| 171 |
+
{"ts": "2026-05-14T15:59:45Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 6998, "success": true}
|
| 172 |
+
{"ts": "2026-05-14T16:00:05Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 31888, "success": true}
|
| 173 |
+
{"ts": "2026-05-14T16:00:06Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 11517, "success": true}
|
| 174 |
+
{"ts": "2026-05-14T16:00:08Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 8264, "success": true}
|
| 175 |
+
{"ts": "2026-05-14T16:00:24Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 38585, "success": true}
|
| 176 |
+
{"ts": "2026-05-14T16:00:25Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5445, "success": true}
|
| 177 |
+
{"ts": "2026-05-14T16:00:29Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": null, "latency_ms": 38423, "success": false}
|
| 178 |
+
{"ts": "2026-05-14T16:00:37Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 13329, "success": true}
|
| 179 |
+
{"ts": "2026-05-14T16:00:42Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 5076, "success": true}
|
| 180 |
+
{"ts": "2026-05-14T16:00:42Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 34690, "success": true}
|
| 181 |
+
{"ts": "2026-05-14T16:00:53Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "nvidia/nemotron-3-nano-30b-a3b", "latency_ms": 23902, "success": true}
|
| 182 |
+
{"ts": "2026-05-14T16:00:53Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 15237, "success": true}
|
| 183 |
+
{"ts": "2026-05-14T16:00:53Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 10166, "success": true}
|
| 184 |
+
{"ts": "2026-05-14T16:00:53Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 845, "success": true}
|
| 185 |
+
{"ts": "2026-05-14T16:00:54Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 956, "success": true}
|
| 186 |
+
{"ts": "2026-05-14T16:00:54Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1043, "success": true}
|
| 187 |
+
{"ts": "2026-05-14T16:01:08Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 15799, "success": true}
|
| 188 |
+
{"ts": "2026-05-14T16:01:14Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "openai/gpt-oss-120b", "latency_ms": 49135, "success": true}
|
| 189 |
+
{"ts": "2026-05-14T16:01:23Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 9747, "success": true}
|
| 190 |
+
{"ts": "2026-05-14T16:01:24Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1017, "success": true}
|
| 191 |
+
{"ts": "2026-05-14T16:01:33Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 13605, "success": true}
|
| 192 |
+
{"ts": "2026-05-14T16:01:33Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": null, "latency_ms": 34443, "success": false}
|
| 193 |
+
{"ts": "2026-05-14T16:01:43Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 10594, "success": true}
|
| 194 |
+
{"ts": "2026-05-14T16:01:53Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 15161, "success": true}
|
| 195 |
+
{"ts": "2026-05-14T16:01:54Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 10494, "success": true}
|
| 196 |
+
{"ts": "2026-05-14T16:01:54Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1045, "success": true}
|
| 197 |
+
{"ts": "2026-05-14T16:02:16Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 11625, "success": true}
|
| 198 |
+
{"ts": "2026-05-14T16:02:17Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1345, "success": true}
|
| 199 |
+
{"ts": "2026-05-14T16:02:24Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 7545, "success": true}
|
| 200 |
+
{"ts": "2026-05-14T16:02:25Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 961, "success": true}
|
| 201 |
+
{"ts": "2026-05-14T16:02:32Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 9210, "success": true}
|
| 202 |
+
{"ts": "2026-05-14T16:02:33Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 972, "success": true}
|
| 203 |
+
{"ts": "2026-05-14T16:02:51Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 10608, "success": true}
|
| 204 |
+
{"ts": "2026-05-14T16:02:52Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1165, "success": true}
|
| 205 |
+
{"ts": "2026-05-14T16:03:05Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 15397, "success": true}
|
| 206 |
+
{"ts": "2026-05-14T16:03:06Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 955, "success": true}
|
| 207 |
+
{"ts": "2026-05-14T16:03:38Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 10987, "success": true}
|
| 208 |
+
{"ts": "2026-05-14T16:03:38Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 893, "success": true}
|
| 209 |
+
{"ts": "2026-05-14T16:04:01Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 9822, "success": true}
|
| 210 |
+
{"ts": "2026-05-14T16:35:23Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "groq:llama-3.3-70b-versatile", "latency_ms": 1938074, "success": true}
|
| 211 |
+
{"ts": "2026-05-14T16:35:33Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": null, "latency_ms": 2013512, "success": false}
|
| 212 |
+
{"ts": "2026-05-14T16:35:34Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 11729, "success": true}
|
| 213 |
+
{"ts": "2026-05-14T16:35:39Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": null, "latency_ms": 1893503, "success": false}
|
| 214 |
+
{"ts": "2026-05-14T16:35:59Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 7438, "success": true}
|
| 215 |
+
{"ts": "2026-05-14T16:35:59Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3188, "success": true}
|
| 216 |
+
{"ts": "2026-05-14T16:36:04Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 16938, "success": true}
|
| 217 |
+
{"ts": "2026-05-14T16:36:07Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3297, "success": true}
|
| 218 |
+
{"ts": "2026-05-14T16:36:17Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 18463, "success": true}
|
| 219 |
+
{"ts": "2026-05-14T16:36:18Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 14559, "success": true}
|
| 220 |
+
{"ts": "2026-05-14T16:36:23Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 16155, "success": true}
|
| 221 |
+
{"ts": "2026-05-14T16:36:24Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 835, "success": true}
|
| 222 |
+
{"ts": "2026-05-14T16:36:25Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1489, "success": true}
|
| 223 |
+
{"ts": "2026-05-14T16:36:40Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1213, "success": true}
|
| 224 |
+
{"ts": "2026-05-14T16:36:42Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 11341, "success": true}
|
| 225 |
+
{"ts": "2026-05-14T16:36:43Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 8840, "success": true}
|
| 226 |
+
{"ts": "2026-05-14T16:36:48Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 8597, "success": true}
|
| 227 |
+
{"ts": "2026-05-14T16:36:50Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 8160, "success": true}
|
| 228 |
+
{"ts": "2026-05-14T16:37:05Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4481, "success": true}
|
| 229 |
+
{"ts": "2026-05-14T16:37:08Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3091, "success": true}
|
| 230 |
+
{"ts": "2026-05-14T16:37:13Z", "role": "paraphraser", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1738, "success": true}
|
| 231 |
+
{"ts": "2026-05-14T16:37:14Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "groq:llama-3.3-70b-versatile", "latency_ms": 30414, "success": true}
|
| 232 |
+
{"ts": "2026-05-14T16:37:16Z", "role": "paraphraser", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2243, "success": true}
|
| 233 |
+
{"ts": "2026-05-14T16:37:16Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 7750, "success": true}
|
| 234 |
+
{"ts": "2026-05-14T16:37:17Z", "role": "paraphraser", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1462, "success": true}
|
| 235 |
+
{"ts": "2026-05-14T16:37:31Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5227, "success": true}
|
| 236 |
+
{"ts": "2026-05-14T16:37:38Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 7711, "success": true}
|
| 237 |
+
{"ts": "2026-05-14T16:37:45Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 6453, "success": true}
|
| 238 |
+
{"ts": "2026-05-14T16:37:51Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 6039, "success": true}
|
| 239 |
+
{"ts": "2026-05-14T16:37:52Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1249, "success": true}
|
| 240 |
+
{"ts": "2026-05-14T16:37:56Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 5142, "success": true}
|
| 241 |
+
{"ts": "2026-05-14T16:38:03Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1051, "success": true}
|
| 242 |
+
{"ts": "2026-05-14T16:38:08Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 4789, "success": true}
|
| 243 |
+
{"ts": "2026-05-14T16:38:21Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1376, "success": true}
|
| 244 |
+
{"ts": "2026-05-14T16:38:22Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1141, "success": true}
|
| 245 |
+
{"ts": "2026-05-14T16:38:23Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "groq:llama-3.3-70b-versatile", "latency_ms": 30442, "success": true}
|
| 246 |
+
{"ts": "2026-05-14T16:38:30Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2013, "success": true}
|
| 247 |
+
{"ts": "2026-05-14T16:38:31Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1694, "success": true}
|
| 248 |
+
{"ts": "2026-05-14T16:38:50Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 9523, "success": true}
|
| 249 |
+
{"ts": "2026-05-14T16:38:51Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1842, "success": true}
|
| 250 |
+
{"ts": "2026-05-14T16:38:55Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 4020, "success": true}
|
| 251 |
+
{"ts": "2026-05-14T16:39:09Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1765, "success": true}
|
| 252 |
+
{"ts": "2026-05-14T16:39:13Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 3903, "success": true}
|
| 253 |
+
{"ts": "2026-05-14T16:39:20Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "groq:llama-3.3-70b-versatile", "latency_ms": 30498, "success": true}
|
| 254 |
+
{"ts": "2026-05-14T16:39:21Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1427, "success": true}
|
| 255 |
+
{"ts": "2026-05-14T16:39:34Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2080, "success": true}
|
| 256 |
+
{"ts": "2026-05-14T16:39:41Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2086, "success": true}
|
| 257 |
+
{"ts": "2026-05-14T16:39:53Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4154, "success": true}
|
| 258 |
+
{"ts": "2026-05-14T16:39:59Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 5041, "success": true}
|
| 259 |
+
{"ts": "2026-05-14T16:40:08Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 9111, "success": true}
|
| 260 |
+
{"ts": "2026-05-14T16:40:09Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 15785, "success": true}
|
| 261 |
+
{"ts": "2026-05-14T16:40:26Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 833, "success": true}
|
| 262 |
+
{"ts": "2026-05-14T16:40:28Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 12805, "success": true}
|
| 263 |
+
{"ts": "2026-05-14T16:40:28Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 8394, "success": true}
|
| 264 |
+
{"ts": "2026-05-14T16:40:30Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1538, "success": true}
|
| 265 |
+
{"ts": "2026-05-14T16:40:38Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2756, "success": true}
|
| 266 |
+
{"ts": "2026-05-14T16:40:40Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 2067, "success": true}
|
| 267 |
+
{"ts": "2026-05-14T16:40:55Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2892, "success": true}
|
| 268 |
+
{"ts": "2026-05-14T16:42:34Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1945, "success": true}
|
| 269 |
+
{"ts": "2026-05-14T16:42:42Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1722, "success": true}
|
| 270 |
+
{"ts": "2026-05-14T16:42:48Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1377, "success": true}
|
| 271 |
+
{"ts": "2026-05-14T16:42:56Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2872, "success": true}
|
| 272 |
+
{"ts": "2026-05-14T16:42:57Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1810, "success": true}
|
| 273 |
+
{"ts": "2026-05-14T16:43:08Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3301, "success": true}
|
| 274 |
+
{"ts": "2026-05-14T16:43:12Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 4189, "success": true}
|
| 275 |
+
{"ts": "2026-05-14T16:43:21Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 3649, "success": true}
|
| 276 |
+
{"ts": "2026-05-14T16:43:23Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1736, "success": true}
|
| 277 |
+
{"ts": "2026-05-14T16:43:34Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1609, "success": true}
|
| 278 |
+
{"ts": "2026-05-14T16:43:35Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1391, "success": true}
|
| 279 |
+
{"ts": "2026-05-14T16:43:44Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "groq:llama-3.3-70b-versatile", "latency_ms": 196617, "success": true}
|
| 280 |
+
{"ts": "2026-05-14T16:43:59Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1338, "success": true}
|
| 281 |
+
{"ts": "2026-05-14T16:44:00Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 6598, "success": true}
|
| 282 |
+
{"ts": "2026-05-14T16:44:03Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 3557, "success": true}
|
| 283 |
+
{"ts": "2026-05-14T16:44:04Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 3372, "success": true}
|
| 284 |
+
{"ts": "2026-05-14T16:44:17Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 7823, "success": true}
|
| 285 |
+
{"ts": "2026-05-14T16:44:18Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1158, "success": true}
|
| 286 |
+
{"ts": "2026-05-14T16:44:19Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2147, "success": true}
|
| 287 |
+
{"ts": "2026-05-14T16:44:26Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 7266, "success": true}
|
| 288 |
+
{"ts": "2026-05-14T16:44:31Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4580, "success": true}
|
| 289 |
+
{"ts": "2026-05-14T16:44:32Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1574, "success": true}
|
| 290 |
+
{"ts": "2026-05-14T16:44:40Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1840, "success": true}
|
| 291 |
+
{"ts": "2026-05-14T16:44:54Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 8136, "success": true}
|
| 292 |
+
{"ts": "2026-05-14T16:44:55Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4891, "success": true}
|
| 293 |
+
{"ts": "2026-05-14T16:44:56Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1425, "success": true}
|
| 294 |
+
{"ts": "2026-05-14T16:44:56Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1099, "success": true}
|
| 295 |
+
{"ts": "2026-05-14T16:45:10Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 4289, "success": true}
|
| 296 |
+
{"ts": "2026-05-14T16:45:11Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1104, "success": true}
|
| 297 |
+
{"ts": "2026-05-14T16:45:12Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1781, "success": true}
|
| 298 |
+
{"ts": "2026-05-14T16:45:13Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1267, "success": true}
|
| 299 |
+
{"ts": "2026-05-14T16:45:29Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 6332, "success": true}
|
| 300 |
+
{"ts": "2026-05-14T16:45:31Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1913, "success": true}
|
| 301 |
+
{"ts": "2026-05-14T16:45:35Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 8083, "success": true}
|
| 302 |
+
{"ts": "2026-05-14T16:45:40Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 5046, "success": true}
|
| 303 |
+
{"ts": "2026-05-14T16:45:43Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 2982, "success": true}
|
| 304 |
+
{"ts": "2026-05-14T16:45:54Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1299, "success": true}
|
| 305 |
+
{"ts": "2026-05-14T16:46:02Z", "role": "fast_brain", "chain_primary": "qwen/qwen3-next-80b-a3b-instruct", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 1862, "success": true}
|
| 306 |
+
{"ts": "2026-05-14T16:53:09Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "groq:llama-3.3-70b-versatile", "latency_ms": 733760, "success": true}
|
| 307 |
+
{"ts": "2026-05-14T17:55:43Z", "role": "profile_extractor", "chain_primary": "nvidia/nemotron-3-nano-30b-a3b", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 633, "success": true}
|
| 308 |
+
{"ts": "2026-05-14T17:55:53Z", "role": "profile_extractor", "chain_primary": "nvidia/nemotron-3-nano-30b-a3b", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 10822, "success": true}
|
| 309 |
+
{"ts": "2026-05-14T17:55:58Z", "role": "profile_extractor", "chain_primary": "nvidia/nemotron-3-nano-30b-a3b", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 15728, "success": true}
|
| 310 |
+
{"ts": "2026-05-14T17:56:03Z", "role": "profile_extractor", "chain_primary": "nvidia/nemotron-3-nano-30b-a3b", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 20393, "success": true}
|
| 311 |
+
{"ts": "2026-05-14T17:56:10Z", "role": "fast_brain", "chain_primary": "groq:llama-3.3-70b-versatile", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 16139, "success": true}
|
| 312 |
+
{"ts": "2026-05-14T17:56:11Z", "role": "judge", "chain_primary": "mistralai/mistral-large-3-675b-instruct-2512", "served_model": "meta/llama-4-maverick-17b-128e-instruct", "latency_ms": 1387, "success": true}
|
| 313 |
+
{"ts": "2026-05-14T17:56:11Z", "role": "profile_extractor", "chain_primary": "nvidia/nemotron-3-nano-30b-a3b", "served_model": "qwen/qwen3-next-80b-a3b-instruct", "latency_ms": 368, "success": true}
|
40-data/premiums/illustrative_premiums.json
CHANGED
|
@@ -16,7 +16,7 @@
|
|
| 16 |
"https://joinditto.in/articles/health-insurance/niva-bupa-reassure-2-0-premium-chart/",
|
| 17 |
"https://joinditto.in/articles/health-insurance/care-supreme-health-insurance-premium-chart/",
|
| 18 |
"https://joinditto.in/articles/health-insurance/tata-aig-health-insurance-premium-rate-chart/",
|
| 19 |
-
"https://www.axisbank.com/docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf",
|
| 20 |
"https://transactions.nivabupa.com/pages/doc/premium_chart/Health%20Premia%20-%20Rate%20Card.pdf",
|
| 21 |
"https://www.scribd.com/document/667873826/Annexure-2-Premium-Chart-Family-Health-Optima-Insurance-Plan-Revised",
|
| 22 |
"https://www.scribd.com/document/682628786/ReAssure-2-0-Rate-Card-with-Tax",
|
|
@@ -762,7 +762,7 @@
|
|
| 762 |
"smoker": false,
|
| 763 |
"family_size": 1,
|
| 764 |
"annual_premium_inr": 7100,
|
| 765 |
-
"source_url": "https://www.axisbank.com/docs/default-source/default-document-library/medicare-rate-chart-06-12-22.pdf",
|
| 766 |
"source_note": "Tata AIG MediCare base variant rate chart - 19-35 / 5L = Rs 7,108 ex-tax"
|
| 767 |
},
|
| 768 |
{
|
|
@@ -772,7 +772,7 @@
|
|
| 772 |
"smoker": false,
|
| 773 |
"family_size": 1,
|
| 774 |
"annual_premium_inr": 8800,
|
| 775 |
-
"source_url": "https://www.axisbank.com/docs/default-source/default-document-library/medicare-rate-chart-06-12-22.pdf"
|
| 776 |
},
|
| 777 |
{
|
| 778 |
"age": 40,
|
|
@@ -796,7 +796,7 @@
|
|
| 796 |
"smoker": false,
|
| 797 |
"family_size": 1,
|
| 798 |
"annual_premium_inr": 8362,
|
| 799 |
-
"source_url": "https://www.axisbank.com/docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf",
|
| 800 |
"source_note": "Official Tata AIG MediCare Premier rate chart: 19-35 age band, 5L SI = Rs 8,362 ex-tax"
|
| 801 |
},
|
| 802 |
{
|
|
@@ -806,7 +806,7 @@
|
|
| 806 |
"smoker": false,
|
| 807 |
"family_size": 1,
|
| 808 |
"annual_premium_inr": 10319,
|
| 809 |
-
"source_url": "https://www.axisbank.com/docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf",
|
| 810 |
"source_note": "Official rate chart: 19-35 / 10L = Rs 10,319 ex-tax"
|
| 811 |
},
|
| 812 |
{
|
|
@@ -816,7 +816,7 @@
|
|
| 816 |
"smoker": false,
|
| 817 |
"family_size": 1,
|
| 818 |
"annual_premium_inr": 9868,
|
| 819 |
-
"source_url": "https://www.axisbank.com/docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf",
|
| 820 |
"source_note": "Official rate chart: 36-45 / 5L = Rs 9,868 ex-tax"
|
| 821 |
},
|
| 822 |
{
|
|
@@ -826,7 +826,7 @@
|
|
| 826 |
"smoker": false,
|
| 827 |
"family_size": 1,
|
| 828 |
"annual_premium_inr": 12266,
|
| 829 |
-
"source_url": "https://www.axisbank.com/docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf",
|
| 830 |
"source_note": "Official rate chart: 36-45 / 10L = Rs 12,266 ex-tax"
|
| 831 |
},
|
| 832 |
{
|
|
|
|
| 16 |
"https://joinditto.in/articles/health-insurance/niva-bupa-reassure-2-0-premium-chart/",
|
| 17 |
"https://joinditto.in/articles/health-insurance/care-supreme-health-insurance-premium-chart/",
|
| 18 |
"https://joinditto.in/articles/health-insurance/tata-aig-health-insurance-premium-rate-chart/",
|
| 19 |
+
"https://www.axisbank.com/70-docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf",
|
| 20 |
"https://transactions.nivabupa.com/pages/doc/premium_chart/Health%20Premia%20-%20Rate%20Card.pdf",
|
| 21 |
"https://www.scribd.com/document/667873826/Annexure-2-Premium-Chart-Family-Health-Optima-Insurance-Plan-Revised",
|
| 22 |
"https://www.scribd.com/document/682628786/ReAssure-2-0-Rate-Card-with-Tax",
|
|
|
|
| 762 |
"smoker": false,
|
| 763 |
"family_size": 1,
|
| 764 |
"annual_premium_inr": 7100,
|
| 765 |
+
"source_url": "https://www.axisbank.com/70-docs/default-source/default-document-library/medicare-rate-chart-06-12-22.pdf",
|
| 766 |
"source_note": "Tata AIG MediCare base variant rate chart - 19-35 / 5L = Rs 7,108 ex-tax"
|
| 767 |
},
|
| 768 |
{
|
|
|
|
| 772 |
"smoker": false,
|
| 773 |
"family_size": 1,
|
| 774 |
"annual_premium_inr": 8800,
|
| 775 |
+
"source_url": "https://www.axisbank.com/70-docs/default-source/default-document-library/medicare-rate-chart-06-12-22.pdf"
|
| 776 |
},
|
| 777 |
{
|
| 778 |
"age": 40,
|
|
|
|
| 796 |
"smoker": false,
|
| 797 |
"family_size": 1,
|
| 798 |
"annual_premium_inr": 8362,
|
| 799 |
+
"source_url": "https://www.axisbank.com/70-docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf",
|
| 800 |
"source_note": "Official Tata AIG MediCare Premier rate chart: 19-35 age band, 5L SI = Rs 8,362 ex-tax"
|
| 801 |
},
|
| 802 |
{
|
|
|
|
| 806 |
"smoker": false,
|
| 807 |
"family_size": 1,
|
| 808 |
"annual_premium_inr": 10319,
|
| 809 |
+
"source_url": "https://www.axisbank.com/70-docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf",
|
| 810 |
"source_note": "Official rate chart: 19-35 / 10L = Rs 10,319 ex-tax"
|
| 811 |
},
|
| 812 |
{
|
|
|
|
| 816 |
"smoker": false,
|
| 817 |
"family_size": 1,
|
| 818 |
"annual_premium_inr": 9868,
|
| 819 |
+
"source_url": "https://www.axisbank.com/70-docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf",
|
| 820 |
"source_note": "Official rate chart: 36-45 / 5L = Rs 9,868 ex-tax"
|
| 821 |
},
|
| 822 |
{
|
|
|
|
| 826 |
"smoker": false,
|
| 827 |
"family_size": 1,
|
| 828 |
"annual_premium_inr": 12266,
|
| 829 |
+
"source_url": "https://www.axisbank.com/70-docs/default-source/default-document-library/tata-aig-medicare-premier-rate-chart.pdf",
|
| 830 |
"source_note": "Official rate chart: 36-45 / 10L = Rs 12,266 ex-tax"
|
| 831 |
},
|
| 832 |
{
|
40-data/profiles/README.md
CHANGED
|
@@ -49,19 +49,19 @@ The original design considered embedding-only. The "why JSON" trade-offs:
|
|
| 49 |
- **Deterministic name lookup.** `Rohit` → `rohit.json` is exact; vector search is approximate and can collide on common first names.
|
| 50 |
- **Human-readable.** A BFSI auditor can `cat` the file and see the full profile.
|
| 51 |
- **Manually editable.** Quick repair without a re-embed pipeline.
|
| 52 |
-
- **No HNSW bloat exposure.** Profile updates do not touch the policy vector store ([ADR-029](../docs/60-decisions/ADR-029-disk-storage-hardening.md)).
|
| 53 |
|
| 54 |
The Chroma chunk is purely a retrieval-time view of the canonical JSON.
|
| 55 |
|
| 56 |
## Privacy + retention
|
| 57 |
|
| 58 |
- Profiles are local to the deployed instance. No third-party share.
|
| 59 |
-
- Per [ADR-010](../docs/60-decisions/ADR-010-secret-handling.md), the folder is not exposed via the HTTP API except through the user's own `session_id`.
|
| 60 |
- The folder is committed empty (placeholder) — actual profiles are runtime artefacts.
|
| 61 |
|
| 62 |
## Related
|
| 63 |
|
| 64 |
- `backend/profile_store.py` — the canonical store implementation
|
| 65 |
-
- `backend/profile_extractor.py` + [ADR-022](../docs/60-decisions/ADR-022-conversational-profile-updates.md) — how conversational asides flow into the profile
|
| 66 |
- `backend/profile_rag.py` — the embedding mirror
|
| 67 |
- `backend/needs_finder.py::GRAPH` — the 9-slot schema the `profile` block conforms to
|
|
|
|
| 49 |
- **Deterministic name lookup.** `Rohit` → `rohit.json` is exact; vector search is approximate and can collide on common first names.
|
| 50 |
- **Human-readable.** A BFSI auditor can `cat` the file and see the full profile.
|
| 51 |
- **Manually editable.** Quick repair without a re-embed pipeline.
|
| 52 |
+
- **No HNSW bloat exposure.** Profile updates do not touch the policy vector store ([ADR-029](../70-docs/60-decisions/ADR-029-disk-storage-hardening.md)).
|
| 53 |
|
| 54 |
The Chroma chunk is purely a retrieval-time view of the canonical JSON.
|
| 55 |
|
| 56 |
## Privacy + retention
|
| 57 |
|
| 58 |
- Profiles are local to the deployed instance. No third-party share.
|
| 59 |
+
- Per [ADR-010](../70-docs/60-decisions/ADR-010-secret-handling.md), the folder is not exposed via the HTTP API except through the user's own `session_id`.
|
| 60 |
- The folder is committed empty (placeholder) — actual profiles are runtime artefacts.
|
| 61 |
|
| 62 |
## Related
|
| 63 |
|
| 64 |
- `backend/profile_store.py` — the canonical store implementation
|
| 65 |
+
- `backend/profile_extractor.py` + [ADR-022](../70-docs/60-decisions/ADR-022-conversational-profile-updates.md) — how conversational asides flow into the profile
|
| 66 |
- `backend/profile_rag.py` — the embedding mirror
|
| 67 |
- `backend/needs_finder.py::GRAPH` — the 9-slot schema the `profile` block conforms to
|
40-data/reviews/icici-lombard.json
CHANGED
|
@@ -171,7 +171,7 @@
|
|
| 171 |
"https://www.business-standard.com/companies/quarterly-results/icici-lombard-net-profit-up-29-in-q1fy26-125071501318_1.html",
|
| 172 |
"https://www.business-standard.com/finance/insurance/icici-lombard-care-health-aditya-birla-cut-commissions-post-gst-125100101265_1.html",
|
| 173 |
"https://www.icicilombard.com/investor-relations",
|
| 174 |
-
"https://www.icicilombard.com/docs/default-source/financial-information/annualreport2024.pdf"
|
| 175 |
]
|
| 176 |
}
|
| 177 |
}
|
|
|
|
| 171 |
"https://www.business-standard.com/companies/quarterly-results/icici-lombard-net-profit-up-29-in-q1fy26-125071501318_1.html",
|
| 172 |
"https://www.business-standard.com/finance/insurance/icici-lombard-care-health-aditya-birla-cut-commissions-post-gst-125100101265_1.html",
|
| 173 |
"https://www.icicilombard.com/investor-relations",
|
| 174 |
+
"https://www.icicilombard.com/70-docs/default-source/financial-information/annualreport2024.pdf"
|
| 175 |
]
|
| 176 |
}
|
| 177 |
}
|
80-audit/ENTERPRISE_AUDIT.md
CHANGED
|
@@ -229,6 +229,19 @@ The full eval was sending **every** QA question to `needs_finder`. Sample bot an
|
|
| 229 |
| KI-032 | `6f495c1` | LLM paraphraser for fact-find questions with verifier + cache (ADR-027) |
|
| 230 |
| KI-033 | `9a977de` | `fact_find_normalizer` + `profile_extractor` migrated from hardcoded single-model to `NimChainLLM(FAST_BRAIN_CHAIN)` |
|
| 231 |
| KI-034/5 | `844ed03` | LRU retrieval cache + `FAST_BRAIN_CHAIN` reordered (Nemotron Nano 30B primary) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 232 |
| D-001 | (multi) | ChromaDB HNSW bloat 3-layer prevention (ADR-029) |
|
| 233 |
| D-002 | (LaunchAgent edit) | Three silently-failing LaunchAgent scripts fixed |
|
| 234 |
| D-009 | `bcb7079` | Removed `tmp_*.py` debug files from repo root |
|
|
@@ -238,11 +251,12 @@ The full eval was sending **every** QA question to `needs_finder`. Sample bot an
|
|
| 238 |
- `tests/test_routing_regression.py` — 15 unit tests, all passing. Pins KI-018 / KI-023 / KI-025 invariants.
|
| 239 |
- 5-Q post-fix smoke (no judge): factual 0% → 60%, nim-chain serving 100% of QA.
|
| 240 |
- Live HF Space smoke (`https://rohitsar567-insurancebot.hf.space`): PED waiting-period question now answers via `nim-chain::nemotron-3-nano-30b-a3b::v4-flash::qa` with a grounded reply, not the old "Happy to help. First, your age?" misroute.
|
| 241 |
-
- **Post-fix parallel 96-Q gold eval (93 of 96 completed; 3 trailing questions killed when the run hung on a NIM rate-limit edge case):**
|
| 242 |
-
- **Factual accuracy: 54.8% (51 / 93)** — up from the pre-fix 41.7% baseline.
|
| 243 |
- **KI-022 JSON-fallback** rescued 7 questions that would have scored 0 on Groq judge JSON errors. Without KI-022 the headline would have been ~47%.
|
| 244 |
- PED waiting-period type — previously 0% pre-fix; samples now: "Bot correctly states the 24‑month waiting period" / "matched_nums=['24']" via regex fallback / "36‑month period and includes source cit…".
|
| 245 |
- Stuck questions: rows 94-96 (all `regulatory_oos` refusals — those routes are already at ~100% earlier in the run; the rate-limit hang affected the brain call, not the refusal logic).
|
|
|
|
| 246 |
- Clean 100-persona audit pending — to run once Batch B (bucket reorg) ships and the HF Space is stable.
|
| 247 |
|
| 248 |
## Pending follow-ups (P1)
|
|
|
|
| 229 |
| KI-032 | `6f495c1` | LLM paraphraser for fact-find questions with verifier + cache (ADR-027) |
|
| 230 |
| KI-033 | `9a977de` | `fact_find_normalizer` + `profile_extractor` migrated from hardcoded single-model to `NimChainLLM(FAST_BRAIN_CHAIN)` |
|
| 231 |
| KI-034/5 | `844ed03` | LRU retrieval cache + `FAST_BRAIN_CHAIN` reordered (Nemotron Nano 30B primary) |
|
| 232 |
+
| KI-036/7/8 | `36ef017` | Greeting flow, strict paraphrase verifier, waiting dots — tightens the first-impression UX surface |
|
| 233 |
+
| KI-039 | `a774b91` | Single Clear-chat button (full reset) — removes ambiguity between "reset" vs "new session" |
|
| 234 |
+
| KI-040 | `4449d44` | Named profiles + Clear-chat that preserves profile — lets users iterate without re-doing fact-find |
|
| 235 |
+
| KI-041/2 | `748ce54` | VAD sensitivity tuning + Live OFF by default + dots on first turn — fixes accidental barge-in on quiet rooms |
|
| 236 |
+
| KI-044 | `86fcc31` | PCM pre-roll via AudioWorklet + 11 folder READMEs — eliminates first-syllable clipping on TTS playback |
|
| 237 |
+
| KI-045 | `930e11e` | Natural-conversation classifier in fact-find: intent_change / off-topic-question escapes fact-find branch into QA. Prevents the bot from droning past a user's pivot. |
|
| 238 |
+
| KI-046 | `28b6114` | Explicit refusal on adversarial/fanciful out-of-corpus questions (space tourism, diamond-tipped surgery). Closes the "absence of exclusion ≠ inclusion" reasoning leak. |
|
| 239 |
+
| KI-047 | `e2d09f9` | Bucket reorg (`docs/` → `70-docs/`, `audit_results/` → `80-audit/`) — Option A safe subset, keeps code dirs untouched |
|
| 240 |
+
| KI-048 | `13a1cf4` | Admin backend: `GET /api/admin/profiles` + `GET /api/admin/performance`, both behind `_check_admin` (IP allowlist + password, 404 on auth fail) |
|
| 241 |
+
| KI-049 | `c8bf1a1` | Retrieval top-k 5 → 10 for table-cell questions (room rent / sub-limit / cap / NCB / etc.). Boosts the chance the policy's structured cap-table chunk lands in context — directly targets the `sub_limit` accuracy gap from D-003. |
|
| 242 |
+
| KI-050 | `52c6351` | Complete `data/` → `40-data/` rename across all Python string-path refs. Finishes KI-047 for runtime code. |
|
| 243 |
+
| KI-051 | `2eae364` | Dockerfile `COPY` paths updated (`data → 40-data`, `docs → 70-docs`). Without this the HF Space rebuild fails on the renamed source dirs. |
|
| 244 |
+
| KI-052 | `c53d167` | Admin panel HTML: 3 lazy-loaded tabs (Profile + Visitor Log, Performance, LLM Chain). Performance pulls KI-048's new endpoint; auth state preserved across tab switches. |
|
| 245 |
| D-001 | (multi) | ChromaDB HNSW bloat 3-layer prevention (ADR-029) |
|
| 246 |
| D-002 | (LaunchAgent edit) | Three silently-failing LaunchAgent scripts fixed |
|
| 247 |
| D-009 | `bcb7079` | Removed `tmp_*.py` debug files from repo root |
|
|
|
|
| 251 |
- `tests/test_routing_regression.py` — 15 unit tests, all passing. Pins KI-018 / KI-023 / KI-025 invariants.
|
| 252 |
- 5-Q post-fix smoke (no judge): factual 0% → 60%, nim-chain serving 100% of QA.
|
| 253 |
- Live HF Space smoke (`https://rohitsar567-insurancebot.hf.space`): PED waiting-period question now answers via `nim-chain::nemotron-3-nano-30b-a3b::v4-flash::qa` with a grounded reply, not the old "Happy to help. First, your age?" misroute.
|
| 254 |
+
- **Post-fix parallel 96-Q gold eval (93 of 96 completed; 3 trailing questions killed when the run hung on a NIM rate-limit edge case) — captured BEFORE KI-046 (refusal precision) and KI-049 (retrieval top-k boost) shipped:**
|
| 255 |
+
- **Factual accuracy: 54.8% (51 / 93)** — up from the pre-fix 41.7% baseline. **This number is now stale; a fresh eval is running to measure the KI-046 + KI-049 lift and the headline will be updated once it lands.**
|
| 256 |
- **KI-022 JSON-fallback** rescued 7 questions that would have scored 0 on Groq judge JSON errors. Without KI-022 the headline would have been ~47%.
|
| 257 |
- PED waiting-period type — previously 0% pre-fix; samples now: "Bot correctly states the 24‑month waiting period" / "matched_nums=['24']" via regex fallback / "36‑month period and includes source cit…".
|
| 258 |
- Stuck questions: rows 94-96 (all `regulatory_oos` refusals — those routes are already at ~100% earlier in the run; the rate-limit hang affected the brain call, not the refusal logic).
|
| 259 |
+
- **Expected directional lift from the two pending fixes:** KI-049 directly targets the `sub_limit` accuracy gap called out under D-003 (structured cap-table chunks now have ~2× chance of landing in context); KI-046 directly targets the `exclusions_oos` refusal-logic gap. Both gaps were the explicit "next bottleneck" in the 54.8% post-mortem.
|
| 260 |
- Clean 100-persona audit pending — to run once Batch B (bucket reorg) ships and the HF Space is stable.
|
| 261 |
|
| 262 |
## Pending follow-ups (P1)
|
CLAUDE.md
CHANGED
|
@@ -36,6 +36,12 @@ Every LLM role is a `NimChainLLM` fallback chain, NOT a hardcoded single model.
|
|
| 36 |
- Paraphrases cached per `(session_id, slot_id)` → max 9 paraphrase calls per session.
|
| 37 |
- Re-asks (`"Sorry, I didn't catch that..."`) skip paraphrase so the user has a stable anchor.
|
| 38 |
- Indic queries use canonical Hindi text (paraphraser is English-only).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
|
| 40 |
## Routing invariants (ADR-N/A — orchestrator.py)
|
| 41 |
|
|
@@ -50,6 +56,23 @@ These are pinned by `tests/test_routing_regression.py`:
|
|
| 50 |
|
| 51 |
`rag/retrieve.py` has an in-process LRU cache keyed by `(query_normalized, top_k, sorted policy_ids, sorted insurer_slugs)`. Cap 256. Cache hit skips both Voyage embed + Chroma query. Invalidates on process restart.
|
| 52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
## Disk + storage hardening (ADR-029)
|
| 54 |
|
| 55 |
Three independent safety layers against ChromaDB HNSW bloat:
|
|
@@ -68,6 +91,10 @@ Three independent safety layers against ChromaDB HNSW bloat:
|
|
| 68 |
- **Data lineage:** `kb/AUDIT_TRAIL.md`.
|
| 69 |
- **Tests:** `tests/test_routing_regression.py` (15 tests pinning routing + load-balance invariants).
|
| 70 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 71 |
## Watch-outs
|
| 72 |
|
| 73 |
- **Never use detached `new Audio()`** — see "Voice UX" above.
|
|
|
|
| 36 |
- Paraphrases cached per `(session_id, slot_id)` → max 9 paraphrase calls per session.
|
| 37 |
- Re-asks (`"Sorry, I didn't catch that..."`) skip paraphrase so the user has a stable anchor.
|
| 38 |
- Indic queries use canonical Hindi text (paraphraser is English-only).
|
| 39 |
+
- **Natural-conversation escape (KI-045):** mid-fact-find, if the user message is intent_change (`"never mind"`, `"actually let me ask"`, `"stop asking"`, etc.) OR off-topic question (ends with `?`, ≥4 words, no slot keywords), `orchestrator.py` exits the fact-find branch, sets `session.free_form_session=True`, and lets the QA path handle it. Prevents the bot from droning past a user's pivot.
|
| 40 |
+
|
| 41 |
+
## Refusal precision (KI-046)
|
| 42 |
+
|
| 43 |
+
- Persona prompt now explicitly instructs the bot to refuse on **fanciful / out-of-scope scenarios** (space tourism, diamond-tipped surgery, fictional procedures) with a specific refusal sentence.
|
| 44 |
+
- Anti-pattern guarded against: "policy doesn't explicitly exclude it → maybe it's covered". This is wrong; absence-of-exclusion is not evidence-of-inclusion.
|
| 45 |
|
| 46 |
## Routing invariants (ADR-N/A — orchestrator.py)
|
| 47 |
|
|
|
|
| 56 |
|
| 57 |
`rag/retrieve.py` has an in-process LRU cache keyed by `(query_normalized, top_k, sorted policy_ids, sorted insurer_slugs)`. Cap 256. Cache hit skips both Voyage embed + Chroma query. Invalidates on process restart.
|
| 58 |
|
| 59 |
+
**Top-k boost for table-cell questions (KI-049):** room rent / sub-limit / cap on / single-private / NCB / co-pay / day-care-limit / etc. triggers bump `top_k` from 5 → 10 for that one query, so the policy's structured cap-table chunk has a higher chance of landing in context. Confined to the trigger query only — does not pollute the cache for downstream non-table queries.
|
| 60 |
+
|
| 61 |
+
## Repo bucket layout (KI-047 / KI-050 / KI-051)
|
| 62 |
+
|
| 63 |
+
Numbered top-level buckets for non-code artifacts (sort lexicographically in `ls`):
|
| 64 |
+
|
| 65 |
+
- `40-data/` ← formerly `data/` — runtime/cached data. All Python string-path refs updated (KI-050). Dockerfile `COPY` paths updated (KI-051).
|
| 66 |
+
- `70-docs/` ← formerly `docs/` — ADRs, design notes, decisions.
|
| 67 |
+
- `80-audit/` ← formerly `audit_results/` — defect register + eval artifacts (this audit lives here).
|
| 68 |
+
|
| 69 |
+
**Code dirs (`backend/`, `frontend/`, `rag/`, `tools/`, `eval/`, `tests/`, `kb/`) kept as-is** — Python forbids leading-digit / hyphen package names, so renaming code dirs would break imports.
|
| 70 |
+
|
| 71 |
+
## Admin panel (KI-048 / KI-052)
|
| 72 |
+
|
| 73 |
+
- **Backend:** `GET /api/admin/profiles` + `GET /api/admin/performance`, both behind `_check_admin` (IP allowlist + `X-Admin-Password` header). Auth failure returns 404 (not 401) so unauthenticated probes can't enumerate endpoints.
|
| 74 |
+
- **Frontend:** admin HTML has 3 lazy-loaded tabs — **Profile + Visitor Log** (pulls `/api/admin/profiles`), **Performance** (pulls `/api/admin/performance`), **LLM Chain** (unchanged from prior). Auth state preserved across tab switches.
|
| 75 |
+
|
| 76 |
## Disk + storage hardening (ADR-029)
|
| 77 |
|
| 78 |
Three independent safety layers against ChromaDB HNSW bloat:
|
|
|
|
| 91 |
- **Data lineage:** `kb/AUDIT_TRAIL.md`.
|
| 92 |
- **Tests:** `tests/test_routing_regression.py` (15 tests pinning routing + load-balance invariants).
|
| 93 |
|
| 94 |
+
## Working-style note (personal memory, not a project decision)
|
| 95 |
+
|
| 96 |
+
**Always parallelize independent work** (per `feedback_always_parallelize.md` in personal memory). On any task touching this project: dispatch agents in parallel when subtasks are independent, batch tool calls in a single message when there are no dependencies. Sequential-by-default wastes wall-clock time.
|
| 97 |
+
|
| 98 |
## Watch-outs
|
| 99 |
|
| 100 |
- **Never use detached `new Audio()`** — see "Voice UX" above.
|
README.md
CHANGED
|
@@ -609,9 +609,9 @@ rag/ingest.py → rag/vectors/chroma.sqlite3 (text chunks + embedd
|
|
| 609 |
│ + HNSW binary files in vectors/ (800-token chunks, 120 overlap)
|
| 610 |
│
|
| 611 |
▼
|
| 612 |
-
data/policy_facts/<policy_id>.json (hand-curated facts for marketplace UI)
|
| 613 |
-
data/premiums/illustrative_premiums.json (anchor table for premium calculator)
|
| 614 |
-
data/reviews/<insurer-slug>.json (IRDAI complaints/10K + sentiment roll-ups)
|
| 615 |
│
|
| 616 |
▼
|
| 617 |
kb/policies/<policy_id>.md (markdown writeup; regen via rag/build_kb.py)
|
|
@@ -623,7 +623,7 @@ Full per-stage detail: [`70-docs/20-data-pipeline/ingestion-policy.md`](70-docs/
|
|
| 623 |
|
| 624 |
| Storage | What lives there | Quota |
|
| 625 |
|---|---|---|
|
| 626 |
-
| HF Space repo (origin) | Code only (`backend/`, `frontend/`, `rag/*.py`, `eval/`, `kb/`, `data/`, configs) | 1 GB free tier |
|
| 627 |
| HF Dataset | Heavy data: `rag/corpus/`, `rag/vectors/`, `rag/extracted/` | 50 GB free tier (we use ~500 MB) |
|
| 628 |
| GitHub code repo | Mirror of HF Space | Standard GitHub |
|
| 629 |
| GitHub data repo (LFS) | Mirror of HF Dataset | 1 GB LFS + 1 GB/mo bandwidth |
|
|
|
|
| 609 |
│ + HNSW binary files in vectors/ (800-token chunks, 120 overlap)
|
| 610 |
│
|
| 611 |
▼
|
| 612 |
+
40-data/policy_facts/<policy_id>.json (hand-curated facts for marketplace UI)
|
| 613 |
+
40-data/premiums/illustrative_premiums.json (anchor table for premium calculator)
|
| 614 |
+
40-data/reviews/<insurer-slug>.json (IRDAI complaints/10K + sentiment roll-ups)
|
| 615 |
│
|
| 616 |
▼
|
| 617 |
kb/policies/<policy_id>.md (markdown writeup; regen via rag/build_kb.py)
|
|
|
|
| 623 |
|
| 624 |
| Storage | What lives there | Quota |
|
| 625 |
|---|---|---|
|
| 626 |
+
| HF Space repo (origin) | Code only (`backend/`, `frontend/`, `rag/*.py`, `eval/`, `kb/`, `40-data/`, configs) | 1 GB free tier |
|
| 627 |
| HF Dataset | Heavy data: `rag/corpus/`, `rag/vectors/`, `rag/extracted/` | 50 GB free tier (we use ~500 MB) |
|
| 628 |
| GitHub code repo | Mirror of HF Space | Standard GitHub |
|
| 629 |
| GitHub data repo (LFS) | Mirror of HF Dataset | 1 GB LFS + 1 GB/mo bandwidth |
|
backend/orchestrator.py
CHANGED
|
@@ -607,10 +607,22 @@ async def handle_turn(
|
|
| 607 |
# "we had a baby", "I was diagnosed with diabetes"). Run a lightweight LLM
|
| 608 |
# extractor, apply high-confidence updates to session.profile, and re-upsert
|
| 609 |
# the profile chunk so THIS turn's retrieval reflects the new state.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 610 |
profile_updates_applied: dict = {}
|
|
|
|
|
|
|
| 611 |
try:
|
| 612 |
-
|
| 613 |
-
|
|
|
|
|
|
|
|
|
|
| 614 |
if extracted:
|
| 615 |
for field_name, new_value in extracted.items():
|
| 616 |
if field_name == "health_conditions":
|
|
|
|
| 607 |
# "we had a baby", "I was diagnosed with diabetes"). Run a lightweight LLM
|
| 608 |
# extractor, apply high-confidence updates to session.profile, and re-upsert
|
| 609 |
# the profile chunk so THIS turn's retrieval reflects the new state.
|
| 610 |
+
#
|
| 611 |
+
# KI-053 (2026-05-14) — eval-mode skip. `eval/run.py --no-extract` sets the
|
| 612 |
+
# INSURANCE_BOT_SKIP_PROFILE_EXTRACTOR env var. Gold-eval questions have
|
| 613 |
+
# empty user_profile + ephemeral sessions so this LLM call wastes a NIM
|
| 614 |
+
# request per question without contributing to grading. Skipping it cuts
|
| 615 |
+
# ~25% off eval wall time at zero quality cost. Production runs (env unset)
|
| 616 |
+
# behaviour is identical to before.
|
| 617 |
profile_updates_applied: dict = {}
|
| 618 |
+
import os as _os_pe
|
| 619 |
+
_skip_extract = _os_pe.environ.get("INSURANCE_BOT_SKIP_PROFILE_EXTRACTOR") == "1"
|
| 620 |
try:
|
| 621 |
+
if _skip_extract:
|
| 622 |
+
extracted = None # eval mode — bypass LLM extractor entirely
|
| 623 |
+
else:
|
| 624 |
+
from backend.profile_extractor import extract_profile_updates
|
| 625 |
+
extracted = await extract_profile_updates(user_text, session.profile)
|
| 626 |
if extracted:
|
| 627 |
for field_name, new_value in extracted.items():
|
| 628 |
if field_name == "health_conditions":
|
eval/gold_qa.json
CHANGED
|
@@ -1120,5 +1120,159 @@
|
|
| 1120 |
"expected_refusal": true,
|
| 1121 |
"language": "en",
|
| 1122 |
"generated_by": "pipeline_c_refusal"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1123 |
}
|
| 1124 |
-
]
|
|
|
|
| 1120 |
"expected_refusal": true,
|
| 1121 |
"language": "en",
|
| 1122 |
"generated_by": "pipeline_c_refusal"
|
| 1123 |
+
},
|
| 1124 |
+
{
|
| 1125 |
+
"id": "tax_planning::section_80d_self_below_60::REFUSE::tax_planning::medium",
|
| 1126 |
+
"policy_id": null,
|
| 1127 |
+
"question": "What is the Section 80D tax deduction limit for a self/family health insurance premium if I am under 60?",
|
| 1128 |
+
"expected_answer": "Bot should refuse or say it cannot advise on Income Tax Act provisions \u2014 the policy corpus does not include Section 80D limits.",
|
| 1129 |
+
"question_type": "tax_planning",
|
| 1130 |
+
"difficulty": "medium",
|
| 1131 |
+
"expected_refusal": true,
|
| 1132 |
+
"language": "en",
|
| 1133 |
+
"generated_by": "pipeline_d_tax"
|
| 1134 |
+
},
|
| 1135 |
+
{
|
| 1136 |
+
"id": "tax_planning::section_80d_senior_parents::REFUSE::tax_planning::medium",
|
| 1137 |
+
"policy_id": null,
|
| 1138 |
+
"question": "If I pay the premium for my parents who are 65 and 68, how much can I claim as a deduction under Section 80D?",
|
| 1139 |
+
"expected_answer": "Bot should refuse or say it cannot advise on Income Tax Act deductions \u2014 80D senior-citizen limits are not stated in any policy document in the corpus.",
|
| 1140 |
+
"question_type": "tax_planning",
|
| 1141 |
+
"difficulty": "hard",
|
| 1142 |
+
"expected_refusal": true,
|
| 1143 |
+
"language": "en",
|
| 1144 |
+
"generated_by": "pipeline_d_tax"
|
| 1145 |
+
},
|
| 1146 |
+
{
|
| 1147 |
+
"id": "tax_planning::section_80ddb_critical_illness::REFUSE::tax_planning::hard",
|
| 1148 |
+
"policy_id": null,
|
| 1149 |
+
"question": "Can I claim a deduction under Section 80DDB for cancer treatment expenses?",
|
| 1150 |
+
"expected_answer": "Bot should refuse \u2014 Section 80DDB rules are part of the Income Tax Act and are not covered in the policy corpus.",
|
| 1151 |
+
"question_type": "tax_planning",
|
| 1152 |
+
"difficulty": "hard",
|
| 1153 |
+
"expected_refusal": true,
|
| 1154 |
+
"language": "en",
|
| 1155 |
+
"generated_by": "pipeline_d_tax"
|
| 1156 |
+
},
|
| 1157 |
+
{
|
| 1158 |
+
"id": "tax_planning::section_80u_disability::REFUSE::tax_planning::hard",
|
| 1159 |
+
"policy_id": null,
|
| 1160 |
+
"question": "I have a 40% disability certificate \u2014 what is my Section 80U deduction?",
|
| 1161 |
+
"expected_answer": "Bot should refuse \u2014 Section 80U personal-disability deductions are an Income Tax Act provision, not a health-policy detail; the corpus does not contain ITA limits.",
|
| 1162 |
+
"question_type": "tax_planning",
|
| 1163 |
+
"difficulty": "hard",
|
| 1164 |
+
"expected_refusal": true,
|
| 1165 |
+
"language": "en",
|
| 1166 |
+
"generated_by": "pipeline_d_tax"
|
| 1167 |
+
},
|
| 1168 |
+
{
|
| 1169 |
+
"id": "tax_planning::gst_rate_health_premium::REFUSE::tax_planning::medium",
|
| 1170 |
+
"policy_id": null,
|
| 1171 |
+
"question": "What is the current GST rate charged on health insurance premiums in India?",
|
| 1172 |
+
"expected_answer": "Bot should refuse or say the policy documents do not state the applicable GST percentage \u2014 GST rates are set by the GST Council, not the insurer's product document.",
|
| 1173 |
+
"question_type": "tax_planning",
|
| 1174 |
+
"difficulty": "medium",
|
| 1175 |
+
"expected_refusal": true,
|
| 1176 |
+
"language": "en",
|
| 1177 |
+
"generated_by": "pipeline_d_tax"
|
| 1178 |
+
},
|
| 1179 |
+
{
|
| 1180 |
+
"id": "tax_planning::topup_plan_tax_savings::REFUSE::tax_planning::medium",
|
| 1181 |
+
"policy_id": null,
|
| 1182 |
+
"question": "If I buy a top-up health plan in addition to my base policy, is the top-up premium also eligible for tax deduction?",
|
| 1183 |
+
"expected_answer": "Bot should refuse on tax-eligibility advice \u2014 deduction eligibility under Section 80D is governed by the Income Tax Act, not the policy document; the bot can only describe policy features.",
|
| 1184 |
+
"question_type": "tax_planning",
|
| 1185 |
+
"difficulty": "medium",
|
| 1186 |
+
"expected_refusal": true,
|
| 1187 |
+
"language": "en",
|
| 1188 |
+
"generated_by": "pipeline_d_tax"
|
| 1189 |
+
},
|
| 1190 |
+
{
|
| 1191 |
+
"id": "tax_planning::section_10_10d_life::REFUSE::tax_planning::hard",
|
| 1192 |
+
"policy_id": null,
|
| 1193 |
+
"question": "Are maturity proceeds from my health insurance policy tax-free under Section 10(10D)?",
|
| 1194 |
+
"expected_answer": "Bot should refuse or clarify that Section 10(10D) applies to life-insurance proceeds and is outside the health-policy corpus; health indemnity policies do not pay maturity proceeds.",
|
| 1195 |
+
"question_type": "tax_planning",
|
| 1196 |
+
"difficulty": "hard",
|
| 1197 |
+
"expected_refusal": true,
|
| 1198 |
+
"language": "en",
|
| 1199 |
+
"generated_by": "pipeline_d_tax"
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"id": "bajaj-allianz__tax-gain__cis::tax_benefit_section::REFUSE::tax_planning::medium",
|
| 1203 |
+
"policy_id": "bajaj-allianz__tax-gain__cis",
|
| 1204 |
+
"question": "The plan is called Tax Gain \u2014 what specific Section 80D tax deduction amount can I claim by buying this policy?",
|
| 1205 |
+
"expected_answer": "Bot should refuse or note that the Tax Gain policy document in the corpus does not state any Section 80D limit \u2014 deduction eligibility and amounts are governed by the Income Tax Act, not the policy document.",
|
| 1206 |
+
"question_type": "tax_planning",
|
| 1207 |
+
"difficulty": "medium",
|
| 1208 |
+
"expected_refusal": true,
|
| 1209 |
+
"language": "en",
|
| 1210 |
+
"generated_by": "pipeline_d_tax"
|
| 1211 |
+
},
|
| 1212 |
+
{
|
| 1213 |
+
"id": "bajaj-allianz__tax-gain__cis::tax_saving_calc::REFUSE::tax_planning::hard",
|
| 1214 |
+
"policy_id": "bajaj-allianz__tax-gain__cis",
|
| 1215 |
+
"question": "If I pay \u20b915,000 premium for Tax Gain at a 30% tax bracket, exactly how much tax will I save in the financial year?",
|
| 1216 |
+
"expected_answer": "Bot should refuse \u2014 the Tax Gain policy document does not contain Section 80D limits or marginal tax-rate guidance; tax-saving calculations require Income Tax Act knowledge that is outside the corpus.",
|
| 1217 |
+
"question_type": "tax_planning",
|
| 1218 |
+
"difficulty": "hard",
|
| 1219 |
+
"expected_refusal": true,
|
| 1220 |
+
"language": "en",
|
| 1221 |
+
"generated_by": "pipeline_d_tax"
|
| 1222 |
+
},
|
| 1223 |
+
{
|
| 1224 |
+
"id": "tax_planning::tds_on_claim_payouts::REFUSE::tax_planning::medium",
|
| 1225 |
+
"policy_id": null,
|
| 1226 |
+
"question": "Will TDS be deducted on the cashless claim amount paid by the insurer to the hospital?",
|
| 1227 |
+
"expected_answer": "Bot should refuse or say TDS treatment of insurance claim payouts is governed by the Income Tax Act and is not addressed in the policy document corpus.",
|
| 1228 |
+
"question_type": "tax_planning",
|
| 1229 |
+
"difficulty": "medium",
|
| 1230 |
+
"expected_refusal": true,
|
| 1231 |
+
"language": "en",
|
| 1232 |
+
"generated_by": "pipeline_d_tax"
|
| 1233 |
+
},
|
| 1234 |
+
{
|
| 1235 |
+
"id": "tax_planning::premium_prepay_tax_year::REFUSE::tax_planning::medium",
|
| 1236 |
+
"policy_id": null,
|
| 1237 |
+
"question": "If I pay a 2-year premium in lump sum in March, can I claim the entire amount as a Section 80D deduction in the same financial year?",
|
| 1238 |
+
"expected_answer": "Bot should refuse \u2014 single-premium vs. multi-year apportionment rules under Section 80D are Income Tax Act provisions; the policy corpus does not include tax-treatment of lump-sum premiums.",
|
| 1239 |
+
"question_type": "tax_planning",
|
| 1240 |
+
"difficulty": "hard",
|
| 1241 |
+
"expected_refusal": true,
|
| 1242 |
+
"language": "en",
|
| 1243 |
+
"generated_by": "pipeline_d_tax"
|
| 1244 |
+
},
|
| 1245 |
+
{
|
| 1246 |
+
"id": "tax_planning::hra_medical_interplay::REFUSE::tax_planning::hard",
|
| 1247 |
+
"policy_id": null,
|
| 1248 |
+
"question": "Can I claim HRA exemption and Section 80D health insurance deduction together in the same year?",
|
| 1249 |
+
"expected_answer": "Bot should refuse \u2014 HRA exemption rules and inter-section deduction interactions are Income Tax Act provisions outside the policy corpus.",
|
| 1250 |
+
"question_type": "tax_planning",
|
| 1251 |
+
"difficulty": "hard",
|
| 1252 |
+
"expected_refusal": true,
|
| 1253 |
+
"language": "en",
|
| 1254 |
+
"generated_by": "pipeline_d_tax"
|
| 1255 |
+
},
|
| 1256 |
+
{
|
| 1257 |
+
"id": "tax_planning::preventive_checkup_80d_sublimit::REFUSE::tax_planning::medium",
|
| 1258 |
+
"policy_id": null,
|
| 1259 |
+
"question": "How much of my preventive health check-up cost can I claim as a deduction under Section 80D, separately from the premium?",
|
| 1260 |
+
"expected_answer": "Bot should refuse on the deduction-amount question \u2014 the policy corpus describes preventive check-up coverage limits, but Section 80D sub-limit for preventive HCU (\u20b95,000 within the overall cap) is an Income Tax Act provision not stated in the documents.",
|
| 1261 |
+
"question_type": "tax_planning",
|
| 1262 |
+
"difficulty": "medium",
|
| 1263 |
+
"expected_refusal": true,
|
| 1264 |
+
"language": "en",
|
| 1265 |
+
"generated_by": "pipeline_d_tax"
|
| 1266 |
+
},
|
| 1267 |
+
{
|
| 1268 |
+
"id": "tax_planning::cash_premium_80d_eligibility::REFUSE::tax_planning::medium",
|
| 1269 |
+
"policy_id": null,
|
| 1270 |
+
"question": "If I pay my health insurance premium in cash, can I still claim the Section 80D deduction?",
|
| 1271 |
+
"expected_answer": "Bot should refuse \u2014 the rule that 80D deductions require non-cash payment (except for preventive check-up) is an Income Tax Act provision; the policy documents in the corpus do not state payment-mode rules for tax eligibility.",
|
| 1272 |
+
"question_type": "tax_planning",
|
| 1273 |
+
"difficulty": "medium",
|
| 1274 |
+
"expected_refusal": true,
|
| 1275 |
+
"language": "en",
|
| 1276 |
+
"generated_by": "pipeline_d_tax"
|
| 1277 |
}
|
| 1278 |
+
]
|
eval/run.py
CHANGED
|
@@ -280,8 +280,19 @@ async def main():
|
|
| 280 |
help="Concurrent questions in flight. KI-024 — was serial, now parallel. "
|
| 281 |
"Cap is NIM's 40 req/min (~2 calls per question); 6 workers gives ~5× "
|
| 282 |
"speedup before saturating. Drop to 1 to reproduce historical serial timing.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 283 |
args = parser.parse_args()
|
| 284 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 285 |
if not GOLD_FILE.exists():
|
| 286 |
print(f"Missing {GOLD_FILE} — run `python -m eval.generate_gold` first")
|
| 287 |
return 1
|
|
|
|
| 280 |
help="Concurrent questions in flight. KI-024 — was serial, now parallel. "
|
| 281 |
"Cap is NIM's 40 req/min (~2 calls per question); 6 workers gives ~5× "
|
| 282 |
"speedup before saturating. Drop to 1 to reproduce historical serial timing.")
|
| 283 |
+
parser.add_argument("--no-extract", action="store_true",
|
| 284 |
+
help="KI-053 — skip backend.profile_extractor LLM call in eval. Eval gold "
|
| 285 |
+
"questions have empty user_profile so the extractor wastes a NIM call "
|
| 286 |
+
"per question. Setting this env var makes orchestrator skip it. "
|
| 287 |
+
"~25%% throughput gain; no impact on grading.")
|
| 288 |
args = parser.parse_args()
|
| 289 |
|
| 290 |
+
# KI-053 — propagate the skip flag via env var so the deep-stack
|
| 291 |
+
# orchestrator code path can read it without API churn.
|
| 292 |
+
if args.no_extract:
|
| 293 |
+
import os as _os
|
| 294 |
+
_os.environ["INSURANCE_BOT_SKIP_PROFILE_EXTRACTOR"] = "1"
|
| 295 |
+
|
| 296 |
if not GOLD_FILE.exists():
|
| 297 |
print(f"Missing {GOLD_FILE} — run `python -m eval.generate_gold` first")
|
| 298 |
return 1
|
kb/AUDIT_TRAIL.md
CHANGED
|
@@ -19,7 +19,7 @@ For a Sarvam interviewer or BFSI compliance auditor: open this file, follow the
|
|
| 19 |
SOURCE PDFs DISCOVERY DOWNLOAD
|
| 20 |
- 10 insurer websites → - corpus_discovery → - rag/download_corpus.py
|
| 21 |
- IRDAI database agent (75 URLs found) - HEAD-check + magic byte
|
| 22 |
-
- regulatory portals - data/corpus_urls.md - retry logic for 403/timeout
|
| 23 |
→ rag/corpus/*.pdf
|
| 24 |
|
| 25 |
[4] [5] [6]
|
|
@@ -58,8 +58,8 @@ RESPONSE TTS UI RENDER
|
|
| 58 |
| Stage | Artifact | Path | Generated by | Verifiable by |
|
| 59 |
| --- | --- | --- | --- | --- |
|
| 60 |
| 1 SOURCE | Raw insurer PDFs | `rag/corpus/<insurer>/*.pdf` | manual + agent crawl | open URL in browser; HEAD-check via `tools/verify_urls.py` |
|
| 61 |
-
| 2 DISCOVERY | URL index | `data/corpus_urls.md` | `Agent` task `a9594655...` (research agent) | git history of discovery agent |
|
| 62 |
-
| 2 DISCOVERY | Regulatory URLs | `data/regulatory_urls.md` | `Agent` task `a87f8380...` | git history |
|
| 63 |
| 3 DOWNLOAD | Manifest | `rag/corpus/_manifest.json` | `rag/download_corpus.py` + `rag/download_retry.py` | re-run download script |
|
| 64 |
| 4 PARSE | per-page text (in-memory) | _ephemeral_ | `rag/ingest.py:read_pdf_pages` | open PDF; pdfplumber on it |
|
| 65 |
| 5 CHUNK | chunks (in-memory) | _ephemeral_ | `rag/ingest.py:chunk_pages` | run ingest with `--dry-run` |
|
|
@@ -109,14 +109,14 @@ Every architectural decision in `70-docs/decisions.md` produces a specific artif
|
|
| 109 |
| Decision | Artifact produced |
|
| 110 |
| --- | --- |
|
| 111 |
| D-001 Vertical slice scope | `kb/policies/` (10 categories only Health) |
|
| 112 |
-
| D-002 Health category | `data/corpus_urls.md` (76 health PDFs) |
|
| 113 |
| D-003 Curated corpus | `rag/corpus/` (no user uploads in v1, expanded in v1.1 with security gates) |
|
| 114 |
| D-004 Hybrid structured + unstructured | `rag/policies.duckdb` + `rag/vectors/` |
|
| 115 |
| D-005 Next.js + FastAPI | `frontend/` + `backend/main.py` |
|
| 116 |
| D-006 Sarvam-first | `backend/providers/sarvam_*.py` |
|
| 117 |
| D-007 Pricing illustrative only | (no `actual_premium` field anywhere) |
|
| 118 |
| D-008 Consultative persona | `backend/persona.py` |
|
| 119 |
-
| D-009 10 insurers / all health policies | `data/corpus_urls.md` (75 URLs) |
|
| 120 |
| D-010 Secret handling | `.env` (chmod 600 + gitignored) |
|
| 121 |
| D-011 Voyage embeddings | _superseded by local BGE; see D-011 revision_ |
|
| 122 |
| D-012 Render hosting | _superseded by HF Spaces_ |
|
|
@@ -179,13 +179,13 @@ Total cost from cold: <$2 (BGE local + ~80 LLM extractions). Total wall-time: ~3
|
|
| 179 |
|
| 180 |
## Batch — 2026-05-14
|
| 181 |
|
| 182 |
-
Three back-to-back curation passes brought the `data/policy_facts/` directory to **102 policies** with verbatim-quote provenance. Mirrored into `kb/policies/` today.
|
| 183 |
|
| 184 |
-
- **Batch 1 — human-research curation (22 policies).** Manual + agent-assisted verbatim extraction from local PDFs in `rag/corpus/` for the 22 highest-priority wordings. Schema: `{value, unit?, source_pdf_path, source_quote}` per field with a `_meta` block (`curated_at`, `primary_source_pdf`, `completeness_pct`, `notes`). Average completeness ≈83.5%. Recorded in [`data/policy_facts/_curation_report.md`](../data/policy_facts/_curation_report.md).
|
| 185 |
- **Batch 2 — regex + pdfplumber pass (43 policies).** Automated pattern extraction across the remaining retail health policy PDFs. Each field carries the same provenance triple; numeric values were validated against the verbatim quote before being written.
|
| 186 |
- **Batch 3 — group / specialty policies (37 policies).** `tools/curate_remaining.py` extended coverage to group, top-up, critical-illness, personal-accident, and specialty riders. Marked with `policy_type` (e.g. `hospital_cash`) where the wording diverged from indemnity templates.
|
| 187 |
|
| 188 |
-
**Verification.** `tools/info_source_map.py` produced [`eval/info_source_map.json`](../eval/info_source_map.json) and [`data/information_source_map.md`](../data/information_source_map.md) with verdict counts: **✅ 798 / ⚠️ 321 /
|
| 189 |
|
| 190 |
**UI / runtime changes shipped today:**
|
| 191 |
|
|
|
|
| 19 |
SOURCE PDFs DISCOVERY DOWNLOAD
|
| 20 |
- 10 insurer websites → - corpus_discovery → - rag/download_corpus.py
|
| 21 |
- IRDAI database agent (75 URLs found) - HEAD-check + magic byte
|
| 22 |
+
- regulatory portals - 40-data/corpus_urls.md - retry logic for 403/timeout
|
| 23 |
→ rag/corpus/*.pdf
|
| 24 |
|
| 25 |
[4] [5] [6]
|
|
|
|
| 58 |
| Stage | Artifact | Path | Generated by | Verifiable by |
|
| 59 |
| --- | --- | --- | --- | --- |
|
| 60 |
| 1 SOURCE | Raw insurer PDFs | `rag/corpus/<insurer>/*.pdf` | manual + agent crawl | open URL in browser; HEAD-check via `tools/verify_urls.py` |
|
| 61 |
+
| 2 DISCOVERY | URL index | `40-data/corpus_urls.md` | `Agent` task `a9594655...` (research agent) | git history of discovery agent |
|
| 62 |
+
| 2 DISCOVERY | Regulatory URLs | `40-data/regulatory_urls.md` | `Agent` task `a87f8380...` | git history |
|
| 63 |
| 3 DOWNLOAD | Manifest | `rag/corpus/_manifest.json` | `rag/download_corpus.py` + `rag/download_retry.py` | re-run download script |
|
| 64 |
| 4 PARSE | per-page text (in-memory) | _ephemeral_ | `rag/ingest.py:read_pdf_pages` | open PDF; pdfplumber on it |
|
| 65 |
| 5 CHUNK | chunks (in-memory) | _ephemeral_ | `rag/ingest.py:chunk_pages` | run ingest with `--dry-run` |
|
|
|
|
| 109 |
| Decision | Artifact produced |
|
| 110 |
| --- | --- |
|
| 111 |
| D-001 Vertical slice scope | `kb/policies/` (10 categories only Health) |
|
| 112 |
+
| D-002 Health category | `40-data/corpus_urls.md` (76 health PDFs) |
|
| 113 |
| D-003 Curated corpus | `rag/corpus/` (no user uploads in v1, expanded in v1.1 with security gates) |
|
| 114 |
| D-004 Hybrid structured + unstructured | `rag/policies.duckdb` + `rag/vectors/` |
|
| 115 |
| D-005 Next.js + FastAPI | `frontend/` + `backend/main.py` |
|
| 116 |
| D-006 Sarvam-first | `backend/providers/sarvam_*.py` |
|
| 117 |
| D-007 Pricing illustrative only | (no `actual_premium` field anywhere) |
|
| 118 |
| D-008 Consultative persona | `backend/persona.py` |
|
| 119 |
+
| D-009 10 insurers / all health policies | `40-data/corpus_urls.md` (75 URLs) |
|
| 120 |
| D-010 Secret handling | `.env` (chmod 600 + gitignored) |
|
| 121 |
| D-011 Voyage embeddings | _superseded by local BGE; see D-011 revision_ |
|
| 122 |
| D-012 Render hosting | _superseded by HF Spaces_ |
|
|
|
|
| 179 |
|
| 180 |
## Batch — 2026-05-14
|
| 181 |
|
| 182 |
+
Three back-to-back curation passes brought the `40-data/policy_facts/` directory to **102 policies** with verbatim-quote provenance. Mirrored into `kb/policies/` today.
|
| 183 |
|
| 184 |
+
- **Batch 1 — human-research curation (22 policies).** Manual + agent-assisted verbatim extraction from local PDFs in `rag/corpus/` for the 22 highest-priority wordings. Schema: `{value, unit?, source_pdf_path, source_quote}` per field with a `_meta` block (`curated_at`, `primary_source_pdf`, `completeness_pct`, `notes`). Average completeness ≈83.5%. Recorded in [`40-data/policy_facts/_curation_report.md`](../40-data/policy_facts/_curation_report.md).
|
| 185 |
- **Batch 2 — regex + pdfplumber pass (43 policies).** Automated pattern extraction across the remaining retail health policy PDFs. Each field carries the same provenance triple; numeric values were validated against the verbatim quote before being written.
|
| 186 |
- **Batch 3 — group / specialty policies (37 policies).** `tools/curate_remaining.py` extended coverage to group, top-up, critical-illness, personal-accident, and specialty riders. Marked with `policy_type` (e.g. `hospital_cash`) where the wording diverged from indemnity templates.
|
| 187 |
|
| 188 |
+
**Verification.** `tools/info_source_map.py` produced [`eval/info_source_map.json`](../eval/info_source_map.json) and [`40-data/information_source_map.md`](../40-data/information_source_map.md) with verdict counts: **✅ 798 / ⚠️ 321 / ❌ 0 / ⏳ 1385.** No ❌ (broken-link) verdicts remain; the ⏳ tail tracks deferred verifications. The ✅:⚠️ ratio is the canonical KPI for source-grounding quality on this dataset.
|
| 189 |
|
| 190 |
**UI / runtime changes shipped today:**
|
| 191 |
|
kb/INDEX.md
CHANGED
|
@@ -303,4 +303,4 @@ kb/
|
|
| 303 |
|
| 304 |
## Provenance convention
|
| 305 |
|
| 306 |
-
Every `policies/<id>.md` file is generated from `data/policy_facts/<id>.json` and preserves the verbatim source quote and source PDF path for every field. JSON is the machine source; markdown is the human-readable mirror. Regenerate the entire kb/ tree by running `.venv/bin/python3 tools/build_kb_mirror.py`.
|
|
|
|
| 303 |
|
| 304 |
## Provenance convention
|
| 305 |
|
| 306 |
+
Every `policies/<id>.md` file is generated from `40-data/policy_facts/<id>.json` and preserves the verbatim source quote and source PDF path for every field. JSON is the machine source; markdown is the human-readable mirror. Regenerate the entire kb/ tree by running `.venv/bin/python3 tools/build_kb_mirror.py`.
|
kb/policies/acko__acko-health-ii__wordings.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/acko__acko-health-ii__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/acko__acko-health-ii__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/acko__acko-health-iii-platinum-lite__wordings.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/acko__acko-health-iii-platinum-lite__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/acko__acko-health-iii-platinum-lite__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/acko__acko-health-iii-platinum__wordings.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/acko__acko-health-iii-platinum__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/acko__acko-health-iii-platinum__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/acko__acko-health-iii__cis.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/acko__acko-health-iii__cis.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/acko__acko-health-iii__cis.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/acko__acko-personal-health__wordings.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/acko__acko-personal-health__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/acko__acko-personal-health__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/acko__arogya-sanjeevani__wordings.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/acko__arogya-sanjeevani__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/acko__arogya-sanjeevani__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/aditya-birla__activ-assure-diamond.md
CHANGED
|
@@ -297,4 +297,4 @@ curated_at: 2026-05-13
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
-
_Mirrored from `data/policy_facts/aditya-birla__activ-assure-diamond.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
+
_Mirrored from `40-data/policy_facts/aditya-birla__activ-assure-diamond.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/aditya-birla__activ-assure-diamond__wordings.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/aditya-birla__activ-assure-diamond__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/aditya-birla__activ-assure-diamond__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/aditya-birla__activ-health-individual__wordings.md
CHANGED
|
@@ -259,4 +259,4 @@ Zone II Zone I 10%
|
|
| 259 |
|
| 260 |
---
|
| 261 |
|
| 262 |
-
_Mirrored from `data/policy_facts/aditya-birla__activ-health-individual__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 259 |
|
| 260 |
---
|
| 261 |
|
| 262 |
+
_Mirrored from `40-data/policy_facts/aditya-birla__activ-health-individual__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/aditya-birla__activ-health.md
CHANGED
|
@@ -297,4 +297,4 @@ curated_at: 2026-05-14
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
-
_Mirrored from `data/policy_facts/aditya-birla__activ-health.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
+
_Mirrored from `40-data/policy_facts/aditya-birla__activ-health.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/aditya-birla__activ-one.md
CHANGED
|
@@ -297,4 +297,4 @@ curated_at: 2026-05-13
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
-
_Mirrored from `data/policy_facts/aditya-birla__activ-one.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
+
_Mirrored from `40-data/policy_facts/aditya-birla__activ-one.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/aditya-birla__activ-one__brochure.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/aditya-birla__activ-one__brochure.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/aditya-birla__activ-one__brochure.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/aditya-birla__activ-secure-cancer-secure__brochure.md
CHANGED
|
@@ -257,4 +257,4 @@ curated_at: 2026-05-14
|
|
| 257 |
|
| 258 |
---
|
| 259 |
|
| 260 |
-
_Mirrored from `data/policy_facts/aditya-birla__activ-secure-cancer-secure__brochure.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 257 |
|
| 258 |
---
|
| 259 |
|
| 260 |
+
_Mirrored from `40-data/policy_facts/aditya-birla__activ-secure-cancer-secure__brochure.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/aditya-birla__activ-secure-personal-accident-cancer-secure__wordings.md
CHANGED
|
@@ -257,4 +257,4 @@ curated_at: 2026-05-14
|
|
| 257 |
|
| 258 |
---
|
| 259 |
|
| 260 |
-
_Mirrored from `data/policy_facts/aditya-birla__activ-secure-personal-accident-cancer-secure__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 257 |
|
| 258 |
---
|
| 259 |
|
| 260 |
+
_Mirrored from `40-data/policy_facts/aditya-birla__activ-secure-personal-accident-cancer-secure__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/aditya-birla__group-activ-health__wordings.md
CHANGED
|
@@ -262,4 +262,4 @@ the Associated Med
|
|
| 262 |
|
| 263 |
---
|
| 264 |
|
| 265 |
-
_Mirrored from `data/policy_facts/aditya-birla__group-activ-health__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 262 |
|
| 263 |
---
|
| 264 |
|
| 265 |
+
_Mirrored from `40-data/policy_facts/aditya-birla__group-activ-health__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__comprehensive-care-plan.md
CHANGED
|
@@ -297,4 +297,4 @@ curated_at: 2026-05-14
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__comprehensive-care-plan.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__comprehensive-care-plan.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__comprehensive-care-plan__wordings.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__comprehensive-care-plan__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__comprehensive-care-plan__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__criti-care__wordings.md
CHANGED
|
@@ -257,4 +257,4 @@ curated_at: 2026-05-14
|
|
| 257 |
|
| 258 |
---
|
| 259 |
|
| 260 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__criti-care__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 257 |
|
| 258 |
---
|
| 259 |
|
| 260 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__criti-care__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__extra-care-plus__wordings.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__extra-care-plus__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__extra-care-plus__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__global-health-care.md
CHANGED
|
@@ -297,4 +297,4 @@ curated_at: 2026-05-14
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__global-health-care.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__global-health-care.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__global-health-care__wordings.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__global-health-care__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__global-health-care__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__group-health-guard-gold__wordings.md
CHANGED
|
@@ -260,4 +260,4 @@ year, then it is agreed that 100%
|
|
| 260 |
|
| 261 |
---
|
| 262 |
|
| 263 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__group-health-guard-gold__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 260 |
|
| 261 |
---
|
| 262 |
|
| 263 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__group-health-guard-gold__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__group-personal-accident__wordings.md
CHANGED
|
@@ -255,4 +255,4 @@ curated_at: 2026-05-14
|
|
| 255 |
|
| 256 |
---
|
| 257 |
|
| 258 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__group-personal-accident__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 255 |
|
| 256 |
---
|
| 257 |
|
| 258 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__group-personal-accident__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__health-guard-gold-individual__wordings.md
CHANGED
|
@@ -261,4 +261,4 @@ then it is agreed that 100%
|
|
| 261 |
|
| 262 |
---
|
| 263 |
|
| 264 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__health-guard-gold-individual__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 261 |
|
| 262 |
---
|
| 263 |
|
| 264 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__health-guard-gold-individual__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__health-guard-gold.md
CHANGED
|
@@ -297,4 +297,4 @@ curated_at: 2026-05-13
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__health-guard-gold.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__health-guard-gold.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__health-guard.md
CHANGED
|
@@ -297,4 +297,4 @@ curated_at: 2026-05-14
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__health-guard.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__health-guard.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__health-guard__wordings.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__health-guard__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__health-guard__wordings.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__silver-health__cis.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__silver-health__cis.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__silver-health__cis.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__tax-gain.md
CHANGED
|
@@ -297,4 +297,4 @@ curated_at: 2026-05-14
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__tax-gain.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__tax-gain.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/bajaj-allianz__tax-gain__cis.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/bajaj-allianz__tax-gain__cis.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/bajaj-allianz__tax-gain__cis.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/care-health__care-advantage-add-ons-protect-plus-care-shield__brochure.md
CHANGED
|
@@ -261,4 +261,4 @@ In-patient care Up to
|
|
| 261 |
|
| 262 |
---
|
| 263 |
|
| 264 |
-
_Mirrored from `data/policy_facts/care-health__care-advantage-add-ons-protect-plus-care-shield__brochure.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 261 |
|
| 262 |
---
|
| 263 |
|
| 264 |
+
_Mirrored from `40-data/policy_facts/care-health__care-advantage-add-ons-protect-plus-care-shield__brochure.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/care-health__care-advantage.md
CHANGED
|
@@ -297,4 +297,4 @@ curated_at: 2026-05-14
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
-
_Mirrored from `data/policy_facts/care-health__care-advantage.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
| 300 |
+
_Mirrored from `40-data/policy_facts/care-health__care-advantage.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
kb/policies/care-health__care-advantage__brochure.md
CHANGED
|
@@ -251,4 +251,4 @@ _no data_
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
-
_Mirrored from `data/policy_facts/care-health__care-advantage__brochure.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|
|
|
|
| 251 |
|
| 252 |
---
|
| 253 |
|
| 254 |
+
_Mirrored from `40-data/policy_facts/care-health__care-advantage__brochure.json`. Provenance — every field's verbatim quote and source PDF path is preserved exactly as curated. Do not hand-edit; regenerate via `tools/build_kb_mirror.py`._
|