Check a bank-detail change before you pay

# Assemble the Decosa bank-detail change check on this machine

You are setting up the Decosa bank-detail change check on this Linux machine for an accounts-payable team or a bookkeeping firm. It returns:
- the email's sender, Reply-To, Return-Path and authentication results against your own vendor file;
- look-alike domains, numbers and addresses not on file, a bank in another country, an account holder who isn't the vendor;
- pressure, secrecy, "don't call" and redirected payments, quoted from the email;
- the call-back plan to the number already on file, and a signed record the call-back and a second approver complete.

Everything is sealed in a signed, hash-chained record anyone can re-check.

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:**
- AP email holds vendor bank details: keep it on this machine; the model route stays local (`direct`), and no domain look-ups are made.
- It never says an email is safe and never releases, blocks or schedules a payment. Every bank-detail change is held for a call-back to the number on file.

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/bank-change-check.zip (4 KB, 8 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 bank-change-check` (the api image carries the same bundle under /app/rehearsal/bank-change-check/;
   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 bank-change-check --bundle bank-change-check.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 bank-detail change is detected", "the look-alike domain is flagged in code", "the Hong Kong account is flagged against the vendor's country"). 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. (The `rules` mode of this tool needs no GPU at all: the `api` service alone runs it.)
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: in the clone, `docker build -f docker/api/Dockerfile -t ${DECOSA_REGISTRY}/decosa-api:0.1.0 .`, and `docker compose build llm` from its compose file. If neither works, stop and tell me.

## 3. Write the compose file

Create `~/decosa-bankchange/.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 records, e.g. Example Agency>"
```

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

```yaml
name: decosa-bankchange
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"
      DECOSA_SESSION_TTL_S: "28800"
      DECOSA_CORS_ORIGIN_REGEX: '^https?://(localhost|127\.0\.0\.1)(:\d+)?$$'
      DECOSA_BANKCHANGE_MAX_CONCURRENT: "6"
    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 (5-10 minutes the first time). `curl -s localhost:8445/healthz` should show `"llm": true`.

## 4. Smoke test

```bash
API=localhost:8445
TOKEN=$(curl -s $API/demo/session -H 'content-type: application/json' -d '{"vertical":"bank-change-check"}' | jq -r .token)
curl -s $API/bank-change/check -H "authorization: Bearer $TOKEN" -H 'content-type: application/json' \
  -d '{"sample_id":"lookalike-urgent"}' > /tmp/out.json
jq '{verdict, change: .change_request.present, signs: [.cues[].cue], callback: .callback.number_on_file}' /tmp/out.json
jq '{record}' /tmp/out.json | curl -s $API/record/verify -H 'content-type: application/json' -d @- | jq '{ok, summary}'
```

The sample `lookalike-urgent` is synthetic. Pass if:
- `.change_request.present` is `true`;
- `.cues[].cue` includes `lookalike_domain` and `bank_country_mismatch`;
- `.callback.number_on_file` is `"+1-312-555-0142"` and `+1-312-555-0199` is under `do_not_use`;
- the verdict never calls the email safe; every receipt has `"status": "attested"` and the record verifies.

Then check that the call-back record refuses the number in the email:

```bash
jq '{record, callback: {caller: "AP clerk", number_dialled: "+1-312-555-0199", spoke_with: "x", outcome: "vendor_confirmed"}, approver: {name: "Controller"}}' /tmp/out.json \
  | curl -s $API/bank-change/signoff -H "authorization: Bearer $TOKEN" -H 'content-type: application/json' -d @- | jq .error
```

Pass if it says the number dialled is not the number on file.

## 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 /bank-change/check` takes `{eml: "<the whole message>", vendors_csv: "<your vendor file as CSV>", mode?: both|rules}` (`rules` needs no GPU). It returns JSON, or streams Server-Sent Events with `Accept: text/event-stream`.
- `GET /bank-change/info` lists what is checked, the rules with their sources and dates, the limits and what is not checked.
- The server stores nothing. Keep each signed record (JSON) with your file. Anyone can re-check it with `POST /record/verify` against the key at `GET /attest/signing-key`. No call leaves the box.
- 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. Hosted routes (off, and leave them off)

`DECOSA_LLM_ROUTE=gateway` would send prompts, which contain your files, to a hosted gateway. Never use it for real files.

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