GitHub hosts HunyuanVideo, Tencent's open-source framework for large-scale video generation models, enabling AI-driven video creation.
Qarinah
FreeStarting price
0Popularity
About Qarinah
Qarinah provides a verifiable memory system for coding agents that preserves project evidence, decisions, and outcomes across sessions. It links symbols, decisions, sessions, and Git worktrees into a searchable graph, delivering bounded cited context packs for subsequent tasks. The tool maintains encrypted archives of exact source snapshots alongside lossy but cited model context packs, ensuring precise recovery or efficient recall without replaying full project history. It supports multiple languages through pinned parsers and deterministic ranking, with crash-recoverable automatic memory via a foreground watcher that journals each phase for safe recovery. The system enforces tamper-evident manifests, verifies ledgers, and provides public-checkout results that reproduce sessions without private data or provider calls. Developers can inspect memories, files, concepts, and relationships through a local graph model and dashboard, while session receipts bind exact lifecycle details without copying event bodies.
Key features
- Verifiable memory graph linking symbols, decisions, and outcomes
- Encrypted archive with content-defined chunking for efficient storage
- Crash-recoverable watcher for automatic memory refresh
- Multi-language support with pinned parsers and deterministic ranking
- Tamper-evident manifests and session receipts
- Local graph model and dashboard for inspecting memories
- Public-checkout verification without provider calls
- Editor integrations for VS Code, Cursor, and JetBrains
Use cases
- Maintaining project context across multiple coding agent sessions
- Recovering exact source snapshots for precise debugging or auditing
- Verifying task completions and outcomes without replaying full history
Pros
- Preserves exact source snapshots and verifiable project evidence
- Supports multiple programming languages with deterministic parsing
- Provides crash-recoverable automatic memory with journaling
- Delivers bounded cited context packs for efficient recall
- Enables public verification of sessions without private data exposure
Cons
- Currently in alpha stage (0.6.0-alpha.1)
- Requires local CLI setup and specific editor integrations
- No explicit mention of cloud or multi-user collaboration features
Frequently asked questions about Qarinah
What does Qarinah do?
Qarinah provides a verifiable memory system for coding agents that preserves project evidence, decisions, and outcomes across sessions. It links symbols, decisions, sessions, and Git worktrees into a searchable graph and delivers bounded cited context packs for subsequent tasks.
Who is Qarinah designed for?
Qarinah is designed for developers and coding agents who need to maintain continuity across sessions without replaying full project history. It supports multi-language projects and integrates with editors like VS Code, Cursor, and JetBrains.
How does Qarinah ensure the accuracy of preserved project data?
Qarinah maintains encrypted archives of exact source snapshots and verifies ledgers to ensure tamper-evident manifests. It also provides public-checkout results that reproduce sessions without private data or provider calls.
Can Qarinah recover interrupted sessions?
Yes, Qarinah includes a crash-recoverable automatic memory system with a foreground watcher that journals each phase, allowing safe recovery of interrupted cycles.
What languages does Qarinah support?
Qarinah supports multiple languages including JavaScript, TypeScript, Python, Go, Rust, Java, Kotlin, C, C++, and C# through pinned parsers and deterministic local ranking.
How do I get started with Qarinah?
To get started, initialize the worktrees using the CLI command `npx qarinah setup` and run the harness or dashboard commands to begin capturing and recalling project context.
