Fix/translation startup

[Ayan-josh-05]

Fix Translation Health Checks and Gateway Request Timeouts

Summary

Fixed two startup and reliability issues found during end-to-end testing of the containerized stack:

1. The translation service's /health endpoint could block for several minutes while Ollama loaded the model.
2. The gateway used a flat 120-second timeout for all backend requests, causing legitimate long-running OCR, translation, and field-mapping requests to fail with a 503.

The fixes make translation health checks non-blocking and introduce realistic, per-service gateway timeouts.

What's Fixed

1. Translation /health no longer blocks on Ollama cold loads

Previously, /health called Ollama synchronously. During a cold model load, this could block for 2+ minutes, causing orchestrators or load balancers polling the endpoint to time out instead of receiving a quick "not ready" response.

Now, a background monitor task checks Ollama independently, while /health simply returns the latest observed state immediately:

• ok — model is loaded and responding
• initializing — Ollama is reachable, but the model is still loading
• unreachable — Ollama is not reachable

The monitor:

• Retries every OLLAMA_HEALTH_RETRY_SECONDS (default: 5s)
• Performs up to OLLAMA_HEALTH_MAX_FAST_RETRIES (default: 12) consecutive fast retries
• Backs off to OLLAMA_HEALTH_BACKOFF_SECONDS (default: 120s) after repeated failures
• Continues retrying indefinitely
• Rechecks a healthy Ollama instance every OLLAMA_HEALTH_RECHECK_SECONDS (default: 30s)

The first background ping also warms up the model automatically during startup.

Commit: 226bdd7

2. Gateway no longer uses a flat 120s timeout

Previously, the gateway used a shared 120-second httpx timeout for all backend requests. This was too short for operations such as:

• Cold Ollama model loads
• Multi-page OCR
• Long-running translation
• Field mapping

The frontend already allows up to 5 minutes for these requests, but the gateway could terminate them first and return a spurious 503 while the backend was still processing.

Added separate, environment-configurable timeouts for each service:

• OCR_REQUEST_TIMEOUT_SECONDS
• TRANSLATION_REQUEST_TIMEOUT_SECONDS
• FIELD_MAPPING_REQUEST_TIMEOUT_SECONDS

All default to 300 seconds (5 minutes) and are passed through _proxy().

Health-check proxy requests retain a separate 5-second timeout, since health checks are now expected to respond immediately.

Commit: b0b1860

How to Test

1. Start the stack

cd lending-poc
docker compose up --build

2. Test translation health

Hit the translation service's /health immediately after startup.

Expected behavior:

• Responds immediately with status: "initializing"
• Eventually changes to status: "ok" once the background monitor successfully pings Ollama

The endpoint should no longer hang during the initial model load.

3. Test long-running gateway requests

After a cold start, trigger an OCR, translation, or field-mapping request through the gateway before Ollama has warmed up.

Expected behavior:

• The request is allowed to run for up to 300 seconds
• The gateway does not prematurely return a 503 at the old 120-second limit
• The request completes successfully if the backend finishes within the configured timeout

4. Optional configuration testing

The following variables can be overridden through .env to verify the tuning behavior:

OLLAMA_HEALTH_RETRY_SECONDS
OLLAMA_HEALTH_MAX_FAST_RETRIES
OLLAMA_HEALTH_BACKOFF_SECONDS
OLLAMA_HEALTH_RECHECK_SECONDS

OCR_REQUEST_TIMEOUT_SECONDS
TRANSLATION_REQUEST_TIMEOUT_SECONDS
FIELD_MAPPING_REQUEST_TIMEOUT_SECONDS

Result

Translation health checks are now fast and non-blocking, while the gateway can accommodate legitimately long-running backend operations without prematurely returning 503 errors.