Hindsight

Long-term memory system for AI agents using biomimetic data structures. Provides three core operations:

• Retain: Store memories — extracts facts, entities, and relationships automatically
• Recall: Retrieve memories via 4 parallel strategies (semantic, BM25, graph, temporal) + reranking
• Reflect: Disposition-aware reasoning using memories and mental models

Services

This template deploys 3 services:

Service	Port	Purpose
PostgreSQL (pgvector)	5432	Database with vector search
Hindsight API	8888	Core memory engine (FastAPI)
Control Plane	9999	Admin dashboard UI (Next.js)

Configuration

1. Set your LLM Provider (openai, anthropic, gemini, groq, etc.)
2. Set your LLM API Key
3. Set your LLM Model (e.g. gpt-4o-mini)
4. Bind domains to the API and Control Plane services

Usage

After deployment, access the API at https://<your-api-domain>/docs for the interactive API documentation, or use the Control Plane UI to manage memory banks.

SDK & Integrations

Hindsight provides SDKs for Python, TypeScript, Go, and Rust, plus integrations with Claude Code, OpenAI Codex, CrewAI, LangGraph, LlamaIndex, and more.

Links

• GitHub
• Documentation
• License: Apache 2.0