Show HN: Flamehaven FileSearch – Self-Hosted RAG Semantic Search(github.com)
github.com
Show HN: Flamehaven FileSearch – Self-Hosted RAG Semantic Search
https://github.com/flamehaven01/Flamehaven-Filesearch
2 comments
We just released Flamehaven FileSearch v1.2.0, and this update focuses heavily on security, scalability, and multi-worker production deployments. The main change is that FileSearch is no longer a “public API”—it now includes a complete authentication and permissions system suitable for enterprise environments.
Key changes:
API Authentication & Authorization All protected endpoints now require Bearer tokens. API keys support fine-grained permissions (upload/search/stores/delete), per-key rate limits, auditing, and SHA256 hashing (plain keys are never stored).
Admin Dashboard A self-contained HTML/JS dashboard for managing API keys, inspecting request statistics, and viewing usage distribution. No external dependencies.
Batch Search API Added a high-throughput endpoint that processes 1–100 queries in either sequential or parallel mode. Includes per-query isolation, priority ordering, and detailed timing metrics.
Redis Cache Backend Optional distributed cache for multi-worker setups. Provides <10ms lookup latency, automatic fallback to local LRU, and ~40–60% reduction in LLM calls.
Deployment Support Includes working examples for Docker, Docker Compose, and Kubernetes (ConfigMap, Secret, Deployment). Meant to be runnable out of the box.
Performance highlights:
Cache hit: <10ms
Cache miss (LLM call): ~0.5–3s
Batch Search (10 queries): ~2–5s
GitHub: https://github.com/flamehaven01/Flamehaven-Filesearch
If anyone has feedback on security design, API ergonomics, or deployment structure, I’d really appreciate it. We’re planning v1.2.1 (admin auth improvements, Redis UI) and v1.3.0 (key rotation + OAuth2/OIDC), so suggestions are welcome.
Key changes:
API Authentication & Authorization All protected endpoints now require Bearer tokens. API keys support fine-grained permissions (upload/search/stores/delete), per-key rate limits, auditing, and SHA256 hashing (plain keys are never stored).
Admin Dashboard A self-contained HTML/JS dashboard for managing API keys, inspecting request statistics, and viewing usage distribution. No external dependencies.
Batch Search API Added a high-throughput endpoint that processes 1–100 queries in either sequential or parallel mode. Includes per-query isolation, priority ordering, and detailed timing metrics.
Redis Cache Backend Optional distributed cache for multi-worker setups. Provides <10ms lookup latency, automatic fallback to local LRU, and ~40–60% reduction in LLM calls.
Deployment Support Includes working examples for Docker, Docker Compose, and Kubernetes (ConfigMap, Secret, Deployment). Meant to be runnable out of the box.
Performance highlights:
Cache hit: <10ms
Cache miss (LLM call): ~0.5–3s
Batch Search (10 queries): ~2–5s
GitHub: https://github.com/flamehaven01/Flamehaven-Filesearch
If anyone has feedback on security design, API ergonomics, or deployment structure, I’d really appreciate it. We’re planning v1.2.1 (admin auth improvements, Redis UI) and v1.3.0 (key rotation + OAuth2/OIDC), so suggestions are welcome.
I released v1.1.0 today.
This update includes a path-traversal fix, FastAPI/Starlette security patches, request-ID tracing, rate limiting, LRU+TTL caching (<10ms hits), and a full set of Prometheus metrics. It’s noticeably more stable than the previous version.
The project is still lightweight and fully self-hostable, which remains a core focus. Thanks!
This update includes a path-traversal fix, FastAPI/Starlette security patches, request-ID tracing, rate limiting, LRU+TTL caching (<10ms hits), and a full set of Prometheus metrics. It’s noticeably more stable than the previous version.
The project is still lightweight and fully self-hostable, which remains a core focus. Thanks!
Key Highlights
5-minute setup — pip install flamehaven-filesearch[api] 100% self-hosted — your documents never leave your infra Python SDK + REST API with FastAPI & Swagger UI SQLite-backed store — simple, portable, and offline-friendly Gemini-powered embeddings with natural language Q&A + citations Docker-ready — works on a $5 VPS or your laptop
Why it matters
Unlike typical RAG stacks that require vector DBs, multiple APIs, and complex infra, Flamehaven FileSearch runs cleanly on Python and SQLite — no dependencies, no vendor lock-in. You own the data, the storage, and the intelligence.
It’s built for teams, researchers, and indie developers who need semantic search they can actually deploy and trust.
Try it
GitHub: https://github.com/flamehaven01/Flamehaven-Filesearch
PyPI: flamehaven-filesearch Docs: /docs after running the API server
#opensource #python #AI #RAG #Gemini #FastAPI #Flamehaven #LLM #developer
Security & Account Notice
Recently, GitHub detected a suspicious login attempt on our organization account. As a precaution, a password reset was enforced. We are currently in contact with GitHub Security and awaiting resolution.
This incident did not affect any public repositories such as dir2md, flashrecord, crom-efficient, ARR-medic-cyp3a4, or Unstable Singularity Detector. All builds remain verified and signed.
We sincerely apologize for any inconvenience caused and appreciate your understanding as we complete the security audit.