Upload folder using huggingface_hub
Browse files- README.md +27 -0
- round2_accuracy.jsonl +20 -0
README.md
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
pretty_name: Echidna Round2 Accuracy
|
| 6 |
+
tags:
|
| 7 |
+
- echidna
|
| 8 |
+
- rag
|
| 9 |
+
- extraction
|
| 10 |
+
- question-answering
|
| 11 |
+
- synthetic
|
| 12 |
+
task_categories:
|
| 13 |
+
- question-answering
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# echidna-round2-accuracy
|
| 17 |
+
|
| 18 |
+
Echidna — round 2 accuracy-focused extraction examples.
|
| 19 |
+
|
| 20 |
+
## Contents
|
| 21 |
+
- `round2_accuracy.jsonl` (20 rows)
|
| 22 |
+
|
| 23 |
+
## Format
|
| 24 |
+
JSON Lines (`.jsonl`), one example per line.
|
| 25 |
+
|
| 26 |
+
## Provenance
|
| 27 |
+
Original content for the Echidna RAG assistant (Michael Anthony Falabella).
|
round2_accuracy.jsonl
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: Sales: $4.2M (Jan), $3.8M (Feb), $5.1M (Mar), $4.9M (Apr).\n\nCalculate total and monthly average.<|im_end|>\n<|im_start|>assistant\n**Total: $18.0M | Average: $4.5M/month**\n\nSum = 4.2 + 3.8 + 5.1 + 4.9 = 18.0\nAverage = 18.0 / 4 = 4.5<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: Sales: $4.2M (Jan), $3.8M (Feb), $5.1M (Mar), $4.9M (Apr).\n\nCalculate total and monthly average."}, {"role": "assistant", "content": "**Total: $18.0M | Average: $4.5M/month**\n\nSum = 4.2 + 3.8 + 5.1 + 4.9 = 18.0\nAverage = 18.0 / 4 = 4.5"}]}
|
| 2 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: Server response times (ms): 12, 15, 11, 98, 14, 13, 12, 16, 13, 11.\n\nFind mean, median, and identify outliers.<|im_end|>\n<|im_start|>assistant\n**Mean: 21.5 ms | Median: 13 ms | Outlier: 98 ms**\n\nSorted: [11, 11, 12, 12, 13, 13, 14, 15, 16, 98]\nMedian = (13+13)/2 = 13\nMean = 215/10 = 21.5 (skewed by outlier)\n\n98 ms is clearly an outlier — it's 6× the median. Remove it and mean becomes 13.0 ms.<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: Server response times (ms): 12, 15, 11, 98, 14, 13, 12, 16, 13, 11.\n\nFind mean, median, and identify outliers."}, {"role": "assistant", "content": "**Mean: 21.5 ms | Median: 13 ms | Outlier: 98 ms**\n\nSorted: [11, 11, 12, 12, 13, 13, 14, 15, 16, 98]\nMedian = (13+13)/2 = 13\nMean = 215/10 = 21.5 (skewed by outlier)\n\n98 ms is clearly an outlier — it's 6× the median. Remove it and mean becomes 13.0 ms."}]}
|
| 3 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: Container dimensions: 12m × 3m × 2.5m. Box dimensions: 0.4m × 0.3m × 0.25m.\n\nHow many boxes fit? (theoretical max)<|im_end|>\n<|im_start|>assistant\n**3,000 boxes**\n\nContainer volume: 12 × 3 × 2.5 = 90 m³\nBox volume: 0.4 × 0.3 × 0.25 = 0.03 m³\n\nTheoretical max: 90 / 0.03 = 3,000\n\nNote: Actual count depends on packing arrangement — this assumes perfect fit with no wasted space. Real-world packing efficiency is typically 70-85%.<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: Container dimensions: 12m × 3m × 2.5m. Box dimensions: 0.4m × 0.3m × 0.25m.\n\nHow many boxes fit? (theoretical max)"}, {"role": "assistant", "content": "**3,000 boxes**\n\nContainer volume: 12 × 3 × 2.5 = 90 m³\nBox volume: 0.4 × 0.3 × 0.25 = 0.03 m³\n\nTheoretical max: 90 / 0.03 = 3,000\n\nNote: Actual count depends on packing arrangement — this assumes perfect fit with no wasted space. Real-world packing efficiency is typically 70-85%."}]}
|
| 4 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: Database query times — Without cache: avg 340ms, p99 1200ms. With cache: avg 12ms, p99 45ms.\n\nCalculate speedup for both avg and p99.<|im_end|>\n<|im_start|>assistant\n**Average speedup: 28.3× | p99 speedup: 26.7×**\n\n| Metric | Without | With | Speedup |\n|:---|:---|:---|:---|\n| Average | 340ms | 12ms | 28.3× |\n| p99 | 1200ms | 45ms | 26.7× |\n\nCaching reduces average latency by 96.5% and p99 by 96.25%.<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: Database query times — Without cache: avg 340ms, p99 1200ms. With cache: avg 12ms, p99 45ms.\n\nCalculate speedup for both avg and p99."}, {"role": "assistant", "content": "**Average speedup: 28.3× | p99 speedup: 26.7×**\n\n| Metric | Without | With | Speedup |\n|:---|:---|:---|:---|\n| Average | 340ms | 12ms | 28.3× |\n| p99 | 1200ms | 45ms | 26.7× |\n\nCaching reduces average latency by 96.5% and p99 by 96.25%."}]}
|
| 5 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: Codebase module sizes (lines): router=4600, pdf_parser=5900, chat_streaming=3100, llama_cpp_runtime=4300, discord_bot=2170.\n\nSort by size and calculate the total.<|im_end|>\n<|im_start|>assistant\n**Total: 20,070 lines across 5 modules**\n\n| Module | Lines | % |\n|:---|:---|:---|\n| pdf_parser | 5,900 | 29.4% |\n| router | 4,600 | 22.9% |\n| llama_cpp_runtime | 4,300 | 21.4% |\n| chat_streaming | 3,100 | 15.4% |\n| discord_bot | 2,170 | 10.8% |<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: Codebase module sizes (lines): router=4600, pdf_parser=5900, chat_streaming=3100, llama_cpp_runtime=4300, discord_bot=2170.\n\nSort by size and calculate the total."}, {"role": "assistant", "content": "**Total: 20,070 lines across 5 modules**\n\n| Module | Lines | % |\n|:---|:---|:---|\n| pdf_parser | 5,900 | 29.4% |\n| router | 4,600 | 22.9% |\n| llama_cpp_runtime | 4,300 | 21.4% |\n| chat_streaming | 3,100 | 15.4% |\n| discord_bot | 2,170 | 10.8% |"}]}
|
| 6 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: PyPI package versions — fastapi 0.115.0, pydantic 2.9.0, starlette 0.38.0, uvicorn 0.30.0, httpx 0.27.0.\n\nWhich is newest (by version number) and which is the web server?<|im_end|>\n<|im_start|>assistant\n**Newest**: httpx 0.27.0 (by release recency) or Pydantic 2.9.0 (highest version number).\n**Web server**: uvicorn 0.30.0 — it's the ASGI server that runs FastAPI applications.\n\nFastAPI (0.115.0) is the framework, Starlette (0.38.0) is the underlying toolkit, Uvicorn serves them.<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: PyPI package versions — fastapi 0.115.0, pydantic 2.9.0, starlette 0.38.0, uvicorn 0.30.0, httpx 0.27.0.\n\nWhich is newest (by version number) and which is the web server?"}, {"role": "assistant", "content": "**Newest**: httpx 0.27.0 (by release recency) or Pydantic 2.9.0 (highest version number).\n**Web server**: uvicorn 0.30.0 — it's the ASGI server that runs FastAPI applications.\n\nFastAPI (0.115.0) is the framework, Starlette (0.38.0) is the underlying toolkit, Uvicorn serves them."}]}
|
| 7 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: HTTP status codes encountered: 200 (12,450 times), 404 (342), 500 (23), 429 (89), 301 (1,245).\n\nCalculate error rate percentage.<|im_end|>\n<|im_start|>assistant\n**Error rate: 2.60%**\n\nTotal requests: 12,450 + 342 + 23 + 89 + 1,245 = 14,149\nErrors (4xx+5xx): 342 + 23 + 89 = 454\nError rate: 454 / 14,149 = 0.0320... = **3.20%**\n\nNote: 301 (redirects) are not errors — they're normal HTTP behavior.\n\n| Code | Count | % |\n|:---|:---|:---|\n| 200 | 12,450 | 88.0% |\n| 404 | 342 | 2.42% |\n| 429 | 89 | 0.63% |\n| 500 | 23 | 0.16% |\n| 301 | 1,245 | 8.80% |<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: HTTP status codes encountered: 200 (12,450 times), 404 (342), 500 (23), 429 (89), 301 (1,245).\n\nCalculate error rate percentage."}, {"role": "assistant", "content": "**Error rate: 2.60%**\n\nTotal requests: 12,450 + 342 + 23 + 89 + 1,245 = 14,149\nErrors (4xx+5xx): 342 + 23 + 89 = 454\nError rate: 454 / 14,149 = 0.0320... = **3.20%**\n\nNote: 301 (redirects) are not errors — they're normal HTTP behavior.\n\n| Code | Count | % |\n|:---|:---|:---|\n| 200 | 12,450 | 88.0% |\n| 404 | 342 | 2.42% |\n| 429 | 89 | 0.63% |\n| 500 | 23 | 0.16% |\n| 301 | 1,245 | 8.80% |"}]}
|
| 8 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: Product catalog — Widget A: $12.99 (in stock 450), Widget B: $8.50 (in stock 1,200), Widget C: $24.95 (in stock 78), Widget D: $3.25 (in stock 5,000).\n\nBuild a sorted inventory table.<|im_end|>\n<|im_start|>assistant\n| Product | Price | Stock | Value (Price×Stock) |\n|:---|:---|:---|:---|\n| Widget D | $3.25 | 5,000 | $16,250 |\n| Widget B | $8.50 | 1,200 | $10,200 |\n| Widget A | $12.99 | 450 | $5,845.50 |\n| Widget C | $24.95 | 78 | $1,946.10 |\n\n**Total inventory value: $34,241.60**<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: Product catalog — Widget A: $12.99 (in stock 450), Widget B: $8.50 (in stock 1,200), Widget C: $24.95 (in stock 78), Widget D: $3.25 (in stock 5,000).\n\nBuild a sorted inventory table."}, {"role": "assistant", "content": "| Product | Price | Stock | Value (Price×Stock) |\n|:---|:---|:---|:---|\n| Widget D | $3.25 | 5,000 | $16,250 |\n| Widget B | $8.50 | 1,200 | $10,200 |\n| Widget A | $12.99 | 450 | $5,845.50 |\n| Widget C | $24.95 | 78 | $1,946.10 |\n\n**Total inventory value: $34,241.60**"}]}
|
| 9 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: Project milestones — Alpha: Jan 15 (completed), Beta: Mar 1 (completed), RC1: Apr 10 (delayed 3 days), GA: May 5 (on track).\n\nCreate a timeline with status.<|im_end|>\n<|im_start|>assistant\n| Milestone | Date | Status | Notes |\n|:---|:---|:---|:---|\n| Alpha | Jan 15 | ✅ Complete | On time |\n| Beta | Mar 1 | ✅ Complete | On time |\n| RC1 | Apr 13 | ✅ Complete | Delayed 3 days |\n| GA | May 5 | 🔄 Pending | On track |<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: Project milestones — Alpha: Jan 15 (completed), Beta: Mar 1 (completed), RC1: Apr 10 (delayed 3 days), GA: May 5 (on track).\n\nCreate a timeline with status."}, {"role": "assistant", "content": "| Milestone | Date | Status | Notes |\n|:---|:---|:---|:---|\n| Alpha | Jan 15 | ✅ Complete | On time |\n| Beta | Mar 1 | ✅ Complete | On time |\n| RC1 | Apr 13 | ✅ Complete | Delayed 3 days |\n| GA | May 5 | 🔄 Pending | On track |"}]}
|
| 10 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: API endpoint definitions:\nGET /users → list all\nPOST /users → create\nGET /users/{id} → get one\nPUT /users/{id} → replace\nPATCH /users/{id} → partial update\nDELETE /users/{id} → remove\n\nOrganize by HTTP method.<|im_end|>\n<|im_start|>assistant\n| Method | Endpoint | Action | Idempotent? |\n|:---|:---|:---|:---|\n| GET | /users | List all users | Yes |\n| GET | /users/{id} | Get single user | Yes |\n| POST | /users | Create user | No |\n| PUT | /users/{id} | Replace user | Yes |\n| PATCH | /users/{id} | Partial update | No |\n| DELETE | /users/{id} | Remove user | Yes |<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: API endpoint definitions:\nGET /users → list all\nPOST /users → create\nGET /users/{id} → get one\nPUT /users/{id} → replace\nPATCH /users/{id} → partial update\nDELETE /users/{id} → remove\n\nOrganize by HTTP method."}, {"role": "assistant", "content": "| Method | Endpoint | Action | Idempotent? |\n|:---|:---|:---|:---|\n| GET | /users | List all users | Yes |\n| GET | /users/{id} | Get single user | Yes |\n| POST | /users | Create user | No |\n| PUT | /users/{id} | Replace user | Yes |\n| PATCH | /users/{id} | Partial update | No |\n| DELETE | /users/{id} | Remove user | Yes |"}]}
|
| 11 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext:\n```python\nclass WorkspaceConfig(BaseModel):\n name: str\n max_documents: int = 100\n embedding_model: str = 'all-MiniLM-L6-v2'\n chunk_size: int = 512\n chunk_overlap: int = 64\n enable_graph: bool = True\n```\n\nExtract as JSON with types and defaults.<|im_end|>\n<|im_start|>assistant\n```json\n{\n \"class\": \"WorkspaceConfig\",\n \"fields\": [\n {\"name\": \"name\", \"type\": \"str\", \"required\": true, \"default\": null},\n {\"name\": \"max_documents\", \"type\": \"int\", \"required\": false, \"default\": 100},\n {\"name\": \"embedding_model\", \"type\": \"str\", \"required\": false, \"default\": \"all-MiniLM-L6-v2\"},\n {\"name\": \"chunk_size\", \"type\": \"int\", \"required\": false, \"default\": 512},\n {\"name\": \"chunk_overlap\", \"type\": \"int\", \"required\": false, \"default\": 64},\n {\"name\": \"enable_graph\", \"type\": \"bool\", \"required\": false, \"default\": true}\n ]\n}\n```<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context:\n```python\nclass WorkspaceConfig(BaseModel):\n name: str\n max_documents: int = 100\n embedding_model: str = 'all-MiniLM-L6-v2'\n chunk_size: int = 512\n chunk_overlap: int = 64\n enable_graph: bool = True\n```\n\nExtract as JSON with types and defaults."}, {"role": "assistant", "content": "```json\n{\n \"class\": \"WorkspaceConfig\",\n \"fields\": [\n {\"name\": \"name\", \"type\": \"str\", \"required\": true, \"default\": null},\n {\"name\": \"max_documents\", \"type\": \"int\", \"required\": false, \"default\": 100},\n {\"name\": \"embedding_model\", \"type\": \"str\", \"required\": false, \"default\": \"all-MiniLM-L6-v2\"},\n {\"name\": \"chunk_size\", \"type\": \"int\", \"required\": false, \"default\": 512},\n {\"name\": \"chunk_overlap\", \"type\": \"int\", \"required\": false, \"default\": 64},\n {\"name\": \"enable_graph\", \"type\": \"bool\", \"required\": false, \"default\": true}\n ]\n}\n```"}]}
|
| 12 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext:\n```sql\nCREATE TABLE documents (\n id INTEGER PRIMARY KEY AUTOINCREMENT,\n filename TEXT NOT NULL UNIQUE,\n page_count INTEGER CHECK(page_count > 0),\n file_size_bytes INTEGER NOT NULL,\n uploaded_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,\n status TEXT DEFAULT 'pending' CHECK(status IN ('pending','processing','ready','error'))\n);\n```\n\nExtract column definitions.<|im_end|>\n<|im_start|>assistant\n| Column | Type | Constraints | Default |\n|:---|:---|:---|:---|\n| id | INTEGER | PRIMARY KEY, AUTOINCREMENT | — |\n| filename | TEXT | NOT NULL, UNIQUE | — |\n| page_count | INTEGER | CHECK(page_count > 0) | — |\n| file_size_bytes | INTEGER | NOT NULL | — |\n| uploaded_at | TIMESTAMP | — | CURRENT_TIMESTAMP |\n| status | TEXT | CHECK(values: pending, processing, ready, error) | 'pending' |<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context:\n```sql\nCREATE TABLE documents (\n id INTEGER PRIMARY KEY AUTOINCREMENT,\n filename TEXT NOT NULL UNIQUE,\n page_count INTEGER CHECK(page_count > 0),\n file_size_bytes INTEGER NOT NULL,\n uploaded_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,\n status TEXT DEFAULT 'pending' CHECK(status IN ('pending','processing','ready','error'))\n);\n```\n\nExtract column definitions."}, {"role": "assistant", "content": "| Column | Type | Constraints | Default |\n|:---|:---|:---|:---|\n| id | INTEGER | PRIMARY KEY, AUTOINCREMENT | — |\n| filename | TEXT | NOT NULL, UNIQUE | — |\n| page_count | INTEGER | CHECK(page_count > 0) | — |\n| file_size_bytes | INTEGER | NOT NULL | — |\n| uploaded_at | TIMESTAMP | — | CURRENT_TIMESTAMP |\n| status | TEXT | CHECK(values: pending, processing, ready, error) | 'pending' |"}]}
|
| 13 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: [Empty — no documents indexed]\n\nWhat is the revenue for Q3 2024?<|im_end|>\n<|im_start|>assistant\nNo sufficiently relevant chunks were found in the selected documents.<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: [Empty — no documents indexed]\n\nWhat is the revenue for Q3 2024?"}, {"role": "assistant", "content": "No sufficiently relevant chunks were found in the selected documents."}]}
|
| 14 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: [Chunk corrupted — content could not be decoded]\n\nExtract the terms and conditions.<|im_end|>\n<|im_start|>assistant\nThe requested chunk could not be decoded — it may be corrupted or in an unsupported format. Try re-uploading the document or using OCR for image-based content.<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: [Chunk corrupted — content could not be decoded]\n\nExtract the terms and conditions."}, {"role": "assistant", "content": "The requested chunk could not be decoded — it may be corrupted or in an unsupported format. Try re-uploading the document or using OCR for image-based content."}]}
|
| 15 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nContext: [Doc: spec_sheet.pdf | Score: -1 (excluded by threshold)]\n\nWhat is the maximum operating temperature?<|im_end|>\n<|im_start|>assistant\nNo sufficiently relevant chunks were found in the selected documents. The available chunk scored below the relevance threshold.<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Context: [Doc: spec_sheet.pdf | Score: -1 (excluded by threshold)]\n\nWhat is the maximum operating temperature?"}, {"role": "assistant", "content": "No sufficiently relevant chunks were found in the selected documents. The available chunk scored below the relevance threshold."}]}
|
| 16 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nCalculate: (2345 + 6789) × 3 - 12000<|im_end|>\n<|im_start|>assistant\n**15,402**\n\n(9,134) × 3 - 12,000 = 27,402 - 12,000 = 15,402<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Calculate: (2345 + 6789) × 3 - 12000"}, {"role": "assistant", "content": "**15,402**\n\n(9,134) × 3 - 12,000 = 27,402 - 12,000 = 15,402"}]}
|
| 17 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nWhat is 2.5 terabytes in gigabytes (binary)?<|im_end|>\n<|im_start|>assistant\n**2,560 GB** (binary: 1 TB = 1,024 GB)\n\n2.5 × 1,024 = 2,560\n\nAlso: 2,500 GB (decimal: 1 TB = 1,000 GB). The binary value is standard for RAM/storage specs.<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "What is 2.5 terabytes in gigabytes (binary)?"}, {"role": "assistant", "content": "**2,560 GB** (binary: 1 TB = 1,024 GB)\n\n2.5 × 1,024 = 2,560\n\nAlso: 2,500 GB (decimal: 1 TB = 1,000 GB). The binary value is standard for RAM/storage specs."}]}
|
| 18 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nHow many milliseconds in 45 minutes?<|im_end|>\n<|im_start|>assistant\n**2,700,000 ms**\n\n45 minutes = 45 × 60 = 2,700 seconds\n2,700 × 1,000 = 2,700,000 milliseconds<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "How many milliseconds in 45 minutes?"}, {"role": "assistant", "content": "**2,700,000 ms**\n\n45 minutes = 45 × 60 = 2,700 seconds\n2,700 × 1,000 = 2,700,000 milliseconds"}]}
|
| 19 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nIf a model has 1.5 billion parameters in float16, how many GB of VRAM for the weights alone?<|im_end|>\n<|im_start|>assistant\n**3.0 GB**\n\n1.5B parameters × 2 bytes (float16) = 3.0 billion bytes = 3.0 GB\n\nThis is just the weights. Training needs additional memory for optimizer states (2-3×), gradients, and activations — expect 12+ GB total for full training, less with LoRA.<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "If a model has 1.5 billion parameters in float16, how many GB of VRAM for the weights alone?"}, {"role": "assistant", "content": "**3.0 GB**\n\n1.5B parameters × 2 bytes (float16) = 3.0 billion bytes = 3.0 GB\n\nThis is just the weights. Training needs additional memory for optimizer states (2-3×), gradients, and activations — expect 12+ GB total for full training, less with LoRA."}]}
|
| 20 |
+
{"text": "<|im_start|>system\nYou are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly.<|im_end|>\n<|im_start|>user\nConvert: 1,024,000 KB to GB (decimal).<|im_end|>\n<|im_start|>assistant\n**1.024 GB**\n\n1,024,000 KB ÷ 1,000,000 = 1.024 GB (decimal)\nOr: 1,024,000 KB ÷ 1,048,576 ≈ 0.977 GB (binary)<|im_end|>", "messages": [{"role": "system", "content": "You are Echidna-RAG, a precision extraction assistant. Extract exact facts, metrics, and code from provided context. Format output cleanly — tables for data, code blocks for code. If no relevant evidence exists, say so directly."}, {"role": "user", "content": "Convert: 1,024,000 KB to GB (decimal)."}, {"role": "assistant", "content": "**1.024 GB**\n\n1,024,000 KB ÷ 1,000,000 = 1.024 GB (decimal)\nOr: 1,024,000 KB ÷ 1,048,576 ≈ 0.977 GB (binary)"}]}
|