Appeal a denial

# Assemble the Decosa claim denial appeal packet on this machine

You are setting up a self-hosted denial appeal packet on this Linux machine for a provider's billing, revenue-cycle or
denials team. It reads a claim denial (EOB, remittance or letter with CARC/RARC codes), the chart excerpts and the payer's
coverage criteria, and returns:
- the next appeal level's filing deadline, with the federal rule and the arithmetic (ERISA and ACA plans, Medicare Parts A
  and B, Medicare Advantage, Medicaid managed care);
- each coverage criterion marked met, not met or undocumented, with the policy quote and a chart quote found word for word;
- a recommendation: appeal, don't appeal on this record, get documentation first, or fix the claim;
- a draft letter, only when the chart supports every criterion, with every sentence checked against the chart and policy;
- a signed, hash-chained packet.

Work step by step, show me each command before running anything that needs sudo, and stop if a check fails.

**Before anything else, remind me:**
- Charts are protected health information. Keep everything on this machine: the model route stays local (`direct`), and
  nothing goes to a hosted service.
- This is a drafting aid. It is not legal or medical advice, it never predicts an outcome, and a person checks the
  criteria, the letter and the deadline against the plan's documents, then files.

Repeat both points in your final summary.

## Step 0: set up with a coding agent, rehearse on mock data, then go private

This prompt is for a coding agent running on the machine that will host the service. We recommend Claude Code with
Claude Opus 5.5; any capable coding agent works. Work in this order:

1. Set up on mock data only. During the whole setup you (the agent) work with the synthetic sample bundle below and
   nothing else. Do not ask me for real data, and do not open, read, list or copy files that hold real data, even to
   "test with something realistic".
2. Rehearse. When the steps below are done and the service is healthy, fetch the mock-data bundle for this tool,
   https://decosa.ai/samples/denial-appeal-packet.zip (6 KB, 10 checks, synthetic or openly licensed: see `licence` in expected.json),
   show me what is in it, and run the rehearsal against the local API:
   `docker compose exec api python scripts/rehearse.py denial-appeal-packet` (the api image carries the same bundle under /app/rehearsal/denial-appeal-packet/;
   with no key set, the script asks the local API for a short demo token). From a decosa-api checkout instead:
   `python scripts/rehearse.py denial-appeal-packet --bundle denial-appeal-packet.zip --base-url http://127.0.0.1:<PORT>`.
   It sends the mock inputs to the local API and prints PASS or FAIL for each expected property (for example: "the redetermination deadline is the notice date + 5 days presumed receipt + 120 days (no model)", "the supported case is recommended for appeal", "the notice date is read from the denial and gives the same deadline"). Show me
   the full output. Every check must pass. If one fails, fix the install and run it again; never edit `expected.json`
   to make a check pass.
3. Stop there. Once the rehearsal passes, tell me, and I will run my own data against the local API myself, on this
   machine.

For the person running this: a coding agent that runs in the cloud sees everything in its context, including files it
reads, command output and anything pasted into the chat. Keep real data out of the chat and out of anything the agent
can read. Switch to your own data only after the rehearsal has passed and the agent's work is done.

## What you are building

| service | image | model | port |
|---|---|---|---|
| `llm` | `${DECOSA_REGISTRY}/decosa-llm:0.1.0` (vLLM 0.29.0, `vllm/vllm-openai@sha256:c2914767605584b6d8f45686b82de173ecc99e781897aa3d0a66dacd72c51ae1`) | `nvidia/Qwen3.8-27B-NVFP4` @ `482ca0f3832238542f8f5295dde86b5f22711d80`, Apache-2.0 | internal 8000 |
| `api` | `${DECOSA_REGISTRY}/decosa-api:0.1.0` (no GPU) | none | `127.0.0.1:8445` |

## 1. Check the GPU, driver and Docker

1. Run `nvidia-smi`. I need one NVIDIA GPU with at least 48 GB and driver 580 or newer.
   - Blackwell (RTX PRO 6000, B200): use the defaults below (NVFP4). This is the measured setup.
   - Hopper (H100/H200) or 48 GB Ada/L40S: set `LLM_MODEL=Qwen/Qwen3.8-27B-FP8` and `LLM_REVISION=main`. On a 48 GB card, also
     set `LLM_MAX_LEN=32768`. Not measured.
   - Under 48 GB: stop and tell me it will not fit.
2. Check `docker --version`, `docker compose version` and `docker run --rm --gpus all ubuntu nvidia-smi`. If Docker or the
   NVIDIA Container Toolkit is missing, install them from the official Docker and NVIDIA repositories. Then run
   `sudo nvidia-ctk runtime configure --runtime=docker` and restart Docker.
3. Confirm about 60 GB of free disk.

## 2. Get the images

The images are **on request** while self-host is in early access: ask at https://decosa.ai/contact?topic=self-host, and Decosa sends the registry (set it as `DECOSA_REGISTRY`), pull access and the compose file. Try `docker pull ${DECOSA_REGISTRY}/decosa-{llm,api}:0.1.0`.

If a pull fails, build from source once the `decosa-api` source is published:
- clone it;
- in the clone, run `docker build -f docker/api/Dockerfile -t ${DECOSA_REGISTRY}/decosa-api:0.1.0 .`;
- run `docker compose build llm` from its compose file.

If neither works, stop and tell me.

## 3. Write the compose file

Create `~/decosa-appeal/.env`:

```bash
DECOSA_TAG=0.1.0
DECOSA_GPU=0
LLM_MODEL=nvidia/Qwen3.8-27B-NVFP4
LLM_REVISION=482ca0f3832238542f8f5295dde86b5f22711d80
LLM_MAX_LEN=65536
LLM_GPU_UTIL=0.85
DECOSA_SIGNER_NAME="<who signs these packets, e.g. Example Clinic revenue cycle>"
```

Create `~/decosa-appeal/docker-compose.yml` with exactly these services:

```yaml
name: decosa-appeal
x-health: &health
  interval: 15s
  timeout: 5s
  retries: 5
services:
  llm:
    image: ${DECOSA_REGISTRY}/decosa-llm:${DECOSA_TAG}
    deploy: { resources: { reservations: { devices: [ { driver: nvidia, device_ids: ["${DECOSA_GPU:-0}"], capabilities: [gpu] } ] } } }
    ipc: host
    restart: unless-stopped
    volumes: [hf-cache:/root/.cache/huggingface]
    command: ["${LLM_MODEL}", "--revision", "${LLM_REVISION}", "--served-model-name", "qwen3.8-27b",
              "--language-model-only", "--max-model-len", "${LLM_MAX_LEN}", "--gpu-memory-utilization", "${LLM_GPU_UTIL}",
              "--max-num-seqs", "16", "--kv-cache-dtype", "fp8_e4m3", "--speculative-config", '{"method":"mtp","num_speculative_tokens":3}',
              "--seed", "0", "--enable-force-include-usage", "--disable-uvicorn-access-log", "--host", "0.0.0.0", "--port", "8000"]
    healthcheck: { <<: *health, test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8000/health', timeout=4)"], start_period: 900s }
  api:
    image: ${DECOSA_REGISTRY}/decosa-api:${DECOSA_TAG}
    restart: unless-stopped
    depends_on: { llm: { condition: service_healthy } }
    environment:
      DECOSA_LLM_ROUTE: direct                  # local model only; receipts are signed by this box's key ("attested")
      DECOSA_LLM_URL: http://llm:8000/v1
      DECOSA_LLM_MODEL: qwen3.8-27b
      DECOSA_LOCAL_SIGNING: "on"                # Ed25519 key created at /data/attest/ed25519.pem on first start
      DECOSA_SIGNER_NAME: ${DECOSA_SIGNER_NAME}
      DECOSA_SESSIONS_PER_IP_HOUR: "1000"
      DECOSA_BUDGET_LLM_TOKENS: "200000"        # per session; one packet needs up to about 10,000 generated tokens
      DECOSA_SESSION_TTL_S: "28800"
      DECOSA_APPEAL_MAX_CONCURRENT: "3"
      DECOSA_CORS_ORIGIN_REGEX: '^https?://(localhost|127\.0\.0\.1)(:\d+)?$$'
    ports: ["127.0.0.1:8445:8445"]
    volumes: [decosa-data:/data]
    healthcheck: { <<: *health, test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8445/healthz', timeout=4)"], start_period: 20s }
volumes: { hf-cache: {}, decosa-data: {} }
```

Use the named volume `decosa-data` exactly as written. A host bind mount owned by root makes the API fail on
`/data/keys.sqlite`.

Run `docker compose up -d`, then poll `docker compose ps` until both services are healthy. The LLM takes 5-10 minutes
the first time. `curl -s localhost:8445/healthz` should show `"llm": true` (`asr` is false here, which is fine).

## 4. Smoke test

The deadline rules need no model:

```bash
API=localhost:8445
curl -s $API/appeal/deadlines -H 'content-type: application/json' \
  -d '{"plan_type":"medicare_ab","level":"initial","notice_date":"2026-09-08"}' | jq '.next | {step, deadline, arithmetic}'
```

Pass if the deadline is `2027-01-11` (5 days' presumed receipt plus 120 days, 42 CFR 405.942).

Then the packets (synthetic cases; the policies are CMS coverage determinations or an invented payer policy):

```bash
TOKEN=$(curl -s $API/demo/session -H 'content-type: application/json' -d '{"vertical":"denial-appeal-packet"}' | jq -r .token)
for s in cpap-appeal cpap-no-appeal admin-missing-npi; do
  curl -s $API/appeal/packet -H "authorization: Bearer $TOKEN" -H 'content-type: application/json' \
    -d "{\"sample_id\":\"$s\"}" > /tmp/appeal-$s.json
  jq -c '{decision: .recommendation.decision, deadline: .deadlines.next.deadline, letter: (.letter != null), receipts: (.receipts|length), statuses: ([.receipts[].status]|unique)}' /tmp/appeal-$s.json
done
jq -r '.criteria[] | "\(.status)\t\(.text)"' /tmp/appeal-cpap-appeal.json
jq '{record}' /tmp/appeal-cpap-appeal.json | curl -s $API/record/verify -H 'content-type: application/json' -d @- | jq '{ok, summary}'
```

Pass if:
- `cpap-appeal` is `appeal` with a letter, deadline `2027-01-11`, and the AHI criterion is met with a quote containing
  "11.2";
- `cpap-no-appeal` is `do_not_appeal` with no letter (the chart says there are no qualifying symptoms);
- `admin-missing-npi` is `fix_claim` with no criteria and no letter;
- every receipt has `"status": "attested"`, and the record verifies (`ok: true`).

## 5. Point the app at the local API

- Base URL: `http://localhost:8445`. For a web app, set `NEXT_PUBLIC_DECOSA_API=http://localhost:8445`. Add other origins to
  `DECOSA_CORS_ORIGINS`.
- `POST /appeal/packet` takes `{denial: {text, plan_type, level?, notice_date?, received_date?, claim_type?, urgent?,
  contract_days?}, records: [{kind?, title?, date?, text}], policy: {title, text}, service?, patient_ref?}`. It returns JSON,
  or streams Server-Sent Events when asked with `Accept: text/event-stream`.
- `GET /appeal/info` lists each deadline rule with its citation, link and the date it was read, the CARC/RARC groups and
  the limits.
- Paste the policy's criteria section, not a URL: nothing is fetched. Give `received_date` when you know it (ERISA and ACA
  count from receipt), and `contract_days` when a provider contract sets a shorter dispute window.
- The server stores nothing. Keep each signed packet (JSON) with the claim. Anyone can re-check it with
  `POST /record/verify` against the key at `GET /attest/signing-key`.
- Keep the API on 127.0.0.1. For other users on the LAN, put a TLS reverse proxy with authentication in front and set
  `DECOSA_TRUSTED_PROXIES`.

## 6. Keep the direct route

`DECOSA_LLM_ROUTE=gateway` would send prompts to the hosted Decosa
API, and those prompts contain the chart and the denial.
Never use it for real patients. At most, use it for synthetic training material.

Finish with a summary: what is running, the health output, the smoke-test results, and the reminders above.