April 29Apr 29 This thread is the support venue for templates published from https://github.com/wisdommen/unraid-templatesHINDSIGHTA self-hosted agent memory system that gives LLM applications persistent, queryable memory with knowledge graph, entity resolution, and multi-strategy semantic retrieval. State-of-the-art on the LongMemEval benchmark.Application Name: hindsightApplication Site: https://hindsight.vectorize.io/Github (project): https://github.com/vectorize-io/hindsightImage Registry: https://github.com/vectorize-io/hindsight/pkgs/container/hindsightLicense: MITDefaultsFully self-hosted: embedded PostgreSQL (pg0), local sentence-transformers for embeddings + reranker, your choice of LLM provider for fact extraction (OpenAI, Anthropic, Gemini, Groq, DeepSeek, MiniMax, Vertex AI, Ollama, or LM Studio)Persistent HuggingFace model cache at /mnt/user/appdata/hindsight/cache (~220 MB after first run)Persistent embedded database at /mnt/user/appdata/hindsight/pg0Endpoints: Web UI on port 9999, REST API + MCP on port 8888OptionalSet HINDSIGHT_API_DATABASE_URL to use external PostgreSQL with the pgvector extension instead of the embedded pg0Set HINDSIGHT_API_LLM_BASE_URL for self-hosted OpenAI-compatible endpoints (Ollama, LM Studio, etc.)First-run notesThe template runs the container as root for the first ~50 ms to chown the bind-mounted cache and database paths to the application user (uid 1000), then drops privileges via runuser. Unraid creates bind paths as 99:100 by default; this wrapper makes the install zero-touch with no manual chown required.First start downloads ~220 MB of model weights (BAAI/bge-small-en-v1.5 + cross-encoder/ms-marco-MiniLM-L-6-v2) into the cache mount. Subsequent restarts are instant.You will need an LLM API key. See https://hindsight.vectorize.io/developer/models for supported providers.SupportReply in this thread with the full container log (Docker tab → Hindsight → Logs) for any startup or runtime issues.For upstream image bugs (the Hindsight server itself), file at https://github.com/vectorize-io/hindsight/issuesFor template / Unraid-specific bugs, this thread or https://github.com/wisdommen/unraid-templates/issues
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.