Text Generation
Transformers
Safetensors
English
qwen3_5
image-text-to-text
kaiju-coder-7
coding
local-ai
business
opencode
tool-use
conversational
Instructions to use RMDWLLC/kaiju-coder-7 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use RMDWLLC/kaiju-coder-7 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="RMDWLLC/kaiju-coder-7") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("RMDWLLC/kaiju-coder-7") model = AutoModelForMultimodalLM.from_pretrained("RMDWLLC/kaiju-coder-7") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use RMDWLLC/kaiju-coder-7 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "RMDWLLC/kaiju-coder-7" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "RMDWLLC/kaiju-coder-7", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/RMDWLLC/kaiju-coder-7
- SGLang
How to use RMDWLLC/kaiju-coder-7 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "RMDWLLC/kaiju-coder-7" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "RMDWLLC/kaiju-coder-7", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "RMDWLLC/kaiju-coder-7" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "RMDWLLC/kaiju-coder-7", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use RMDWLLC/kaiju-coder-7 with Docker Model Runner:
docker model run hf.co/RMDWLLC/kaiju-coder-7
Refresh Kaiju Coder 7 paid API readiness evidence
Browse files- FINAL_RELEASE_REPORT.md +10 -12
- GOAL_COMPLETION_AUDIT.md +4 -4
- PAID_API_READINESS.md +44 -9
FINAL_RELEASE_REPORT.md
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
# Kaiju Coder 7 Final Release Report
|
| 2 |
|
| 3 |
-
Generated: `2026-06-
|
| 4 |
|
| 5 |
Product name: `Kaiju Coder 7`
|
| 6 |
Public model id: `kaiju-coder-7`
|
|
@@ -15,9 +15,10 @@ commercial API launch yet. The local model path, OpenCode profile, harnessed
|
|
| 15 |
business-owner evals, staged helper repos, runtime-quantized recipe, and paid
|
| 16 |
API scaffold are in place. The adapter, OpenCode helper, runtime helper, and
|
| 17 |
merged 53.8GB model are uploaded under `RMDWLLC` with public visibility. The
|
| 18 |
-
paid API has human approval
|
| 19 |
-
|
| 20 |
-
|
|
|
|
| 21 |
|
| 22 |
## Runtime
|
| 23 |
|
|
@@ -41,7 +42,7 @@ stability and speed.
|
|
| 41 |
| Hugging Face release readiness | `ready=True pass=23 fail=0 manual=1 rc=0` |
|
| 42 |
| Public launch readiness | `ready=False pass=23 fail=1 manual=0 rc=1` |
|
| 43 |
| Hugging Face staging integrity | `ready=True pass=6 fail=0 manual=0 rc=0` |
|
| 44 |
-
| Paid API launch readiness | `ready=False pass=
|
| 45 |
|
| 46 |
## Hugging Face Upload Evidence
|
| 47 |
|
|
@@ -60,24 +61,21 @@ stability and speed.
|
|
| 60 |
|
| 61 |
| Status | Check | Detail |
|
| 62 |
|---|---|---|
|
| 63 |
-
| manual | paid API launch preflight |
|
| 64 |
|
| 65 |
## Public Launch Blockers
|
| 66 |
|
| 67 |
| Status | Check | Detail |
|
| 68 |
|---|---|---|
|
| 69 |
-
| fail | paid API launch preflight |
|
| 70 |
|
| 71 |
## Paid API Launch Blockers
|
| 72 |
|
| 73 |
| Status | Check | Detail |
|
| 74 |
|---|---|---|
|
| 75 |
-
| manual | public route mode |
|
| 76 |
-
| manual | wrangler secret list confirms KAIJU_ORIGIN_URL, KAIJU_ORIGIN_SECRET, and KAIJU_STRIPE_WEBHOOK_SECRET |
|
| 77 |
| manual | Stripe Checkout top-up products and webhook endpoint tested with metadata.kaiju_api_key_id | attach sanitized evidence in /Users/richardecholsai7/Apps/kaiju-coder/release/paid-api-launch-evidence.json key `stripe_checkout_topup_staging` |
|
| 78 |
-
| manual | staging request passed through Worker to Gojira-B origin with model=kaiju-coder-7 | attach sanitized evidence in /Users/richardecholsai7/Apps/kaiju-coder/release/paid-api-launch-evidence.json key `worker_to_gojira_staging_request` |
|
| 79 |
-
| manual | rollback command or route switch was exercised and recorded | attach sanitized evidence in /Users/richardecholsai7/Apps/kaiju-coder/release/paid-api-launch-evidence.json key `rollback_exercised` |
|
| 80 |
-
| manual | p95 latency for paid routes is recorded after staging traffic | attach sanitized evidence in /Users/richardecholsai7/Apps/kaiju-coder/release/paid-api-launch-evidence.json key `paid_route_latency` |
|
| 81 |
|
| 82 |
## Evidence Paths
|
| 83 |
|
|
|
|
| 1 |
# Kaiju Coder 7 Final Release Report
|
| 2 |
|
| 3 |
+
Generated: `2026-06-03T21:00:45Z`
|
| 4 |
|
| 5 |
Product name: `Kaiju Coder 7`
|
| 6 |
Public model id: `kaiju-coder-7`
|
|
|
|
| 15 |
business-owner evals, staged helper repos, runtime-quantized recipe, and paid
|
| 16 |
API scaffold are in place. The adapter, OpenCode helper, runtime helper, and
|
| 17 |
merged 53.8GB model are uploaded under `RMDWLLC` with public visibility. The
|
| 18 |
+
paid API has human approval and a live Worker staging route with Cloudflare
|
| 19 |
+
bindings, Worker-to-Gojira proof, rollback proof, and latency evidence. Public
|
| 20 |
+
charging still requires a custom API domain plus real Stripe webhook and
|
| 21 |
+
Checkout top-up evidence.
|
| 22 |
|
| 23 |
## Runtime
|
| 24 |
|
|
|
|
| 42 |
| Hugging Face release readiness | `ready=True pass=23 fail=0 manual=1 rc=0` |
|
| 43 |
| Public launch readiness | `ready=False pass=23 fail=1 manual=0 rc=1` |
|
| 44 |
| Hugging Face staging integrity | `ready=True pass=6 fail=0 manual=0 rc=0` |
|
| 45 |
+
| Paid API launch readiness | `ready=False pass=24 fail=0 manual=3 rc=1` |
|
| 46 |
|
| 47 |
## Hugging Face Upload Evidence
|
| 48 |
|
|
|
|
| 61 |
|
| 62 |
| Status | Check | Detail |
|
| 63 |
|---|---|---|
|
| 64 |
+
| manual | paid API launch preflight | 24 pass, 0 fail, 3 manual |
|
| 65 |
|
| 66 |
## Public Launch Blockers
|
| 67 |
|
| 68 |
| Status | Check | Detail |
|
| 69 |
|---|---|---|
|
| 70 |
+
| fail | paid API launch preflight | 24 pass, 0 fail, 3 manual |
|
| 71 |
|
| 72 |
## Paid API Launch Blockers
|
| 73 |
|
| 74 |
| Status | Check | Detail |
|
| 75 |
|---|---|---|
|
| 76 |
+
| manual | public route mode | public route evidence must use exposure_mode=custom_domain before paid launch |
|
| 77 |
+
| manual | wrangler secret list confirms KAIJU_ORIGIN_URL, KAIJU_ORIGIN_SECRET, and KAIJU_STRIPE_WEBHOOK_SECRET | secret-name evidence missing: KAIJU_STRIPE_WEBHOOK_SECRET |
|
| 78 |
| manual | Stripe Checkout top-up products and webhook endpoint tested with metadata.kaiju_api_key_id | attach sanitized evidence in /Users/richardecholsai7/Apps/kaiju-coder/release/paid-api-launch-evidence.json key `stripe_checkout_topup_staging` |
|
|
|
|
|
|
|
|
|
|
| 79 |
|
| 80 |
## Evidence Paths
|
| 81 |
|
GOAL_COMPLETION_AUDIT.md
CHANGED
|
@@ -1,11 +1,11 @@
|
|
| 1 |
# Kaiju Coder 7 Goal Completion Audit
|
| 2 |
|
| 3 |
-
Generated: `2026-06-
|
| 4 |
|
| 5 |
Overall: `not complete`
|
| 6 |
Summary: `17 passed / 1 blocked / 0 manual`
|
| 7 |
|
| 8 |
-
This audit maps the active Kaiju Coder 7 objective to current evidence. It is stricter than local readiness: local public testing
|
| 9 |
|
| 10 |
## Readiness Commands
|
| 11 |
|
|
@@ -40,12 +40,12 @@ This audit maps the active Kaiju Coder 7 objective to current evidence. It is st
|
|
| 40 |
| Quality | Model/harness prompts produce file-oriented business-owner artifacts rather than vague advice. | `passed` | kaiju_harness/business_suite.py; release/EVAL_SCOREBOARD.md | |
|
| 41 |
| Provenance | Training/eval provenance is preserved and public docs avoid internal checkpoint naming except license/provenance attribution. | `passed` | release/SOURCE_INVENTORY.md; release/DATA_PROVENANCE_DRAFT.md; release/PUBLIC_TESTING_QUICKSTART.md | |
|
| 42 |
| Paid API | Paid API scaffold covers API keys, Stripe billing, rate limits, logging controls, abuse controls, rollback plan, and pricing assumptions. | `passed` | python3 scripts/check_paid_api_readiness.py --mode scaffold; gateway/cloudflare-worker tests | |
|
| 43 |
-
| Paid API | Paid API is ready for public charging. | `blocked` | python3 scripts/check_paid_api_readiness.py --mode launch | Requires
|
| 44 |
| Final Report | Final report includes exact commands run, eval results, changed files, remaining risks, and what Richard should test first. | `passed` | release/FINAL_RELEASE_REPORT.md | |
|
| 45 |
|
| 46 |
## Blocking Items
|
| 47 |
|
| 48 |
-
- Paid API: Paid API is ready for public charging.: Requires
|
| 49 |
|
| 50 |
## Commands To Re-run
|
| 51 |
|
|
|
|
| 1 |
# Kaiju Coder 7 Goal Completion Audit
|
| 2 |
|
| 3 |
+
Generated: `2026-06-03T21:00:45Z`
|
| 4 |
|
| 5 |
Overall: `not complete`
|
| 6 |
Summary: `17 passed / 1 blocked / 0 manual`
|
| 7 |
|
| 8 |
+
This audit maps the active Kaiju Coder 7 objective to current evidence. It is stricter than local readiness: local public testing and Hugging Face release checks can pass while paid API launch remains blocked.
|
| 9 |
|
| 10 |
## Readiness Commands
|
| 11 |
|
|
|
|
| 40 |
| Quality | Model/harness prompts produce file-oriented business-owner artifacts rather than vague advice. | `passed` | kaiju_harness/business_suite.py; release/EVAL_SCOREBOARD.md | |
|
| 41 |
| Provenance | Training/eval provenance is preserved and public docs avoid internal checkpoint naming except license/provenance attribution. | `passed` | release/SOURCE_INVENTORY.md; release/DATA_PROVENANCE_DRAFT.md; release/PUBLIC_TESTING_QUICKSTART.md | |
|
| 42 |
| Paid API | Paid API scaffold covers API keys, Stripe billing, rate limits, logging controls, abuse controls, rollback plan, and pricing assumptions. | `passed` | python3 scripts/check_paid_api_readiness.py --mode scaffold; gateway/cloudflare-worker tests | |
|
| 43 |
+
| Paid API | Paid API is ready for public charging. | `blocked` | python3 scripts/check_paid_api_readiness.py --mode launch | Requires a custom API domain, KAIJU_STRIPE_WEBHOOK_SECRET, and real Stripe Checkout top-up staging evidence. |
|
| 44 |
| Final Report | Final report includes exact commands run, eval results, changed files, remaining risks, and what Richard should test first. | `passed` | release/FINAL_RELEASE_REPORT.md | |
|
| 45 |
|
| 46 |
## Blocking Items
|
| 47 |
|
| 48 |
+
- Paid API: Paid API is ready for public charging.: Requires a custom API domain, KAIJU_STRIPE_WEBHOOK_SECRET, and real Stripe Checkout top-up staging evidence.
|
| 49 |
|
| 50 |
## Commands To Re-run
|
| 51 |
|
PAID_API_READINESS.md
CHANGED
|
@@ -64,15 +64,45 @@ Live Cloudflare resources were created on 2026-06-03 after Wrangler login:
|
|
| 64 |
- R2 `kaiju-api-artifacts` bound as `KAIJU_ARTIFACT_BUCKET`
|
| 65 |
- D1 migration `0001_paid_api.sql` applied successfully
|
| 66 |
|
| 67 |
-
|
| 68 |
|
| 69 |
```text
|
| 70 |
-
|
| 71 |
```
|
| 72 |
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 76 |
|
| 77 |
Covered locally:
|
| 78 |
|
|
@@ -187,9 +217,11 @@ python3 scripts/apply_paid_api_cloudflare_bindings.py \
|
|
| 187 |
```
|
| 188 |
|
| 189 |
`--mode scaffold` verifies the local gateway implementation and should pass.
|
| 190 |
-
`--mode launch` is stricter
|
| 191 |
-
|
| 192 |
-
|
|
|
|
|
|
|
| 193 |
|
| 194 |
Launch evidence is attached through a sanitized JSON file:
|
| 195 |
|
|
@@ -257,7 +289,10 @@ Current quality evidence:
|
|
| 257 |
|
| 258 |
## Release Blockers
|
| 259 |
|
| 260 |
-
-
|
|
|
|
|
|
|
|
|
|
| 261 |
- Harnessed customer-readiness route passes; paid API must route through that
|
| 262 |
deterministic product path until a faster raw/quantized path passes.
|
| 263 |
- Context-size benchmarks passed at 12k, 16k, 24k, and 32k, but the current
|
|
|
|
| 64 |
- R2 `kaiju-api-artifacts` bound as `KAIJU_ARTIFACT_BUCKET`
|
| 65 |
- D1 migration `0001_paid_api.sql` applied successfully
|
| 66 |
|
| 67 |
+
The Worker was deployed on 2026-06-03 at:
|
| 68 |
|
| 69 |
```text
|
| 70 |
+
https://kaiju-api-gateway.kiyomi-api.workers.dev
|
| 71 |
```
|
| 72 |
|
| 73 |
+
Gojira-B now advertises `kaiju-coder-7` from its public health endpoint. The
|
| 74 |
+
origin secret was rotated during launch verification and re-applied to
|
| 75 |
+
Cloudflare without writing the value to this repo.
|
| 76 |
+
|
| 77 |
+
Current launch preflight after live staging traffic and rollback verification:
|
| 78 |
+
|
| 79 |
+
```text
|
| 80 |
+
24 pass / 0 fail / 3 manual
|
| 81 |
+
```
|
| 82 |
+
|
| 83 |
+
Passed live launch evidence:
|
| 84 |
+
|
| 85 |
+
- `workers.dev` public staging route resolves to the Kaiju Worker.
|
| 86 |
+
- `KAIJU_ORIGIN_URL` and `KAIJU_ORIGIN_SECRET` are present by Wrangler secret
|
| 87 |
+
name; `KAIJU_STRIPE_WEBHOOK_SECRET` is intentionally not marked present until
|
| 88 |
+
a real Stripe webhook signing secret is configured.
|
| 89 |
+
- Worker-to-Gojira staging request passed through `/v1/chat/completions` with
|
| 90 |
+
`model=kaiju-coder-7`, HTTP `200`, and streaming enabled.
|
| 91 |
+
- Paid-route latency was measured over five staging samples with p95
|
| 92 |
+
`17686.74ms`.
|
| 93 |
+
- Rollback drill succeeded by deploying same-code version
|
| 94 |
+
`e838e01d-2d72-4eb7-9814-b95b7e2cef14`, rolling traffic back to verified
|
| 95 |
+
version `d37d60d1-7bfc-4ac9-a69c-e9339b5e495f`, and rechecking `/health`.
|
| 96 |
+
|
| 97 |
+
Remaining paid-launch manual items:
|
| 98 |
+
|
| 99 |
+
- Attach a real custom API domain. The current route is public and usable for
|
| 100 |
+
staging, but the launch checker requires `exposure_mode=custom_domain`.
|
| 101 |
+
`rmdw.ai` currently uses Vercel DNS, so a Cloudflare custom domain requires a
|
| 102 |
+
Cloudflare-managed zone, partial/CNAME setup, or a separate API domain.
|
| 103 |
+
- Configure `KAIJU_STRIPE_WEBHOOK_SECRET` from a real Stripe webhook endpoint.
|
| 104 |
+
- Test a real Stripe Checkout top-up that sends `checkout.session.completed`
|
| 105 |
+
with `metadata.kaiju_api_key_id`, including duplicate webhook idempotency.
|
| 106 |
|
| 107 |
Covered locally:
|
| 108 |
|
|
|
|
| 217 |
```
|
| 218 |
|
| 219 |
`--mode scaffold` verifies the local gateway implementation and should pass.
|
| 220 |
+
`--mode launch` is stricter. It should remain red until a real custom API
|
| 221 |
+
domain, Stripe webhook signing secret, and Stripe Checkout top-up staging
|
| 222 |
+
evidence are attached. Live Cloudflare bindings, Worker-to-Gojira staging
|
| 223 |
+
traffic, paid-route latency evidence, and rollback proof are already recorded
|
| 224 |
+
in `release/paid-api-launch-evidence.json`.
|
| 225 |
|
| 226 |
Launch evidence is attached through a sanitized JSON file:
|
| 227 |
|
|
|
|
| 289 |
|
| 290 |
## Release Blockers
|
| 291 |
|
| 292 |
+
- Paid API launch still needs a custom API domain and real Stripe
|
| 293 |
+
Checkout/webhook evidence before public charging.
|
| 294 |
+
- Raw OpenCode customer-readiness task currently times out on multi-file work;
|
| 295 |
+
the harnessed business-owner route is the reliable first paid API product.
|
| 296 |
- Harnessed customer-readiness route passes; paid API must route through that
|
| 297 |
deterministic product path until a faster raw/quantized path passes.
|
| 298 |
- Context-size benchmarks passed at 12k, 16k, 24k, and 32k, but the current
|