AI assistant context
These pages are canonical project context for Codex (AGENTS.md), Cursor (rules + optional AGENTS.md), Claude Code (CLAUDE.md imports), and Gemini CLI (GEMINI.md imports).
| Doc | Purpose |
|---|---|
| Project map | Directories, entry points, generated IL files. |
| PDF pipeline | TRANSLATE_STAGES and stage boundaries. |
| Translator stack | OpenAI vs router vs local; config merge pitfalls. |
| Verification | Commands and “done means”. |
| Benchmarks (human docs) | Perf harness: tests/perf/, meso scripts, Locust, nightly workflows. |
| Package layers | Optional extras, public APIs, CI profiles. |
| Docker (human docs) | Container images, volumes, warm vs slim targets (profiles). |
| Library API (human docs) | Stable Python/JSON embedding (doctranslate.api / doctranslate.schemas). |
| HTTP API (human docs) | Optional FastAPI service (doctranslate serve, DocTranslater[api]). |
| HTTP API Python client (human docs) | Generated doc_translater_http_api_client from OpenAPI + convenience helpers. |
| Observability (human docs) | Logs, Prometheus /metrics, request IDs, optional OTLP tracing. |
| HTTP API storage (human docs) | SQLite metadata, fsspec/S3/GCS mirrors, TTL, migration. |
| Serverless containers (human docs) | Cloud Run / Fargate / App Runner / Modal / Runpod patterns and image matrix. |
| Serverless runtime reference (human docs) | Env vars and image ↔ workload compatibility for deployments. |
| Public API policy (human docs) | Semver boundaries and internal vs public modules. |
| MCP suggestions | Which MCPs help this repo. |
| Phase 2 skills | Optional follow-up skills/agents. |
Human contributors can ignore this folder; it does not change runtime behavior.