sulabhdubey /
rta-smriti-brain
Local project memory for AI coding agents, with evidence labels on every claim
- 94 stars
- 3 stars since 3 Oct
- Python
- Updated 4 days ago
- Advanced
- Memory, RAG & data
Every new AI coding session starts by asking what is true about the project right now. Code is in the repository, decisions are in old chats, and test results are in tools. Rta-Smriti pulls these into one local database so an agent can pick up where the last one stopped.
It indexes the Git checkout into project-scoped SQLite, recording file hashes, symbols, imports, evidence and memories. It keeps kinds of fact apart: a test result or a human instruction counts as accepted truth, while an inference or brainstorm needs review before promotion. When a task starts, a context compiler picks direct evidence first, respects a token budget, and writes a receipt of what it included.
It is an evidence and continuity layer, not an agent. It does not run project work or choose models. It offers a CLI, a local MCP server and a loopback web console. Encrypted sharing of selected memory between devices is optional and off by default. The README calls the current release an advanced early-adopter alpha.
What it does
Evidence-labelled memory
Claims keep their source, hash, freshness and verification state. Observations and trusted instructions are kept apart from inferences and hypotheses, which need review.
Bounded context packs
The context compiler selects direct evidence first, follows token budgets and privacy grants, and emits a receipt. It abstains if the named project anchor is missing.
Repository graph
Indexes files, symbols, imports, calls, tests, memories and evidence so you can explore a project beyond keyword search.
Session capture and continuation
Agent sessions are captured incrementally with redaction, and structured checkpoints let a fresh session continue the work.
Works with several agent hosts
A local stdio MCP server plus recipes for Codex, Claude Code, Cursor, Zed, OpenCode and Gemini CLI. The README notes recipes are not all tested live.
What you could build
- Resume a coding task in a fresh agent session without re-explaining decisions and constraints.
- Check whether a recalled fact about a project is backed by a test result or only a guess.
- Explore a repository's files, symbols, calls and tests as a graph in a local console.
- Give Claude Code or Codex a token-limited context pack through MCP.
- Share selected project memory with a teammate through encrypted, scoped sync.
Get started
You need Python 3.11 or newer and Git. Create a virtual environment, install with pip install ., then run the rta-brain start command on your project folder. It indexes the repo and opens a local console. The README has PowerShell and macOS/Linux versions, plus standalone binaries.
git clone https://github.com/sulabhdubey/rta-smriti-brain.gitcd rta-smriti-brainpython3 -m venv .venvCommands are copied from the project's README. Read it on GitHub for the full setup.
More in Memory, RAG & data
firecrawl
Web API that turns sites and search results into markdown or JSON for LLMs
markitdown
Python tool that turns PDFs, Office files, audio and more into Markdown for LLMs
graphify
Maps a codebase and its docs into a knowledge graph your AI assistant can query
claude-mem
Saves what your coding agent does and feeds the useful parts back into later sessions
vectorize-io/hindsight
Memory server that lets AI agents store, search and learn from past interactions
OpenViking
A context database that stores agent memory, docs and skills as browsable files
open-notebook
Self-hosted research notebook that chats with your documents using any AI provider
WeKnora
Self-hosted platform that turns documents into RAG search, an agent and a wiki
supermemory
Memory layer that lets AI assistants and apps remember users across conversations
gbrain
A shared memory for AI agents that stores sourced facts and answers with citations