# Ontave

> Open-source TypeScript framework for source-linked evidence, reviewed assertions, additive ontology evolution and typed accepted-data queries.

Current public npm release: 0.1.0-alpha.4. Developer preview. Node.js 24, ESM and PostgreSQL 18 on a trusted server. The host supplies authentication, authorization, worker supervision, model configuration and review policy.

Ingestion preserves evidence and a durable job. Interpretation produces proposals. Model activation and fact acceptance are separate explicit decisions. Accepted records can be queried through a pinned contract without an LLM call at read time. Acceptance does not establish independent verification of a reported event.

Ontave does not supply an agent orchestrator, hosted database, browser/edge runtime, MCP server, automatic app generation or autonomous fact acceptance. Independent labels, live identity/discovery qualification and other MVP quality gates remain open.

## Start here
- [Documentation](https://ontave.daltn.tech/docs/): Concepts and integration guides.
- [Install from npm](https://www.npmjs.com/package/ontave): npm install ontave.
- [Quickstart](https://ontave.daltn.tech/docs/quickstart/): Fictional local playground using the real SDK and PostgreSQL.
- [API reference](https://ontave.daltn.tech/docs/api/): Configuration, operations and constraints.
- [Release status](https://ontave.daltn.tech/docs/status/): Supported surface and open quality gates.

## Architecture and use cases
- [How to build an AI agent in TypeScript](https://ontave.daltn.tech/index/build-ai-agent-typescript/index.md): A read-only agent tutorial with a bounded tool loop, fictional records and offline tests.
- [Choosing the best AI agent framework for your application](https://ontave.daltn.tech/index/best-ai-agent-frameworks/index.md): Compare AI SDK, LangGraph, Mastra, Pydantic AI and CrewAI by the work each controls.
- [Agent memory vs. a database: what belongs where?](https://ontave.daltn.tech/index/agent-memory-vs-database/index.md): Separate conversation context, execution state, retrieved passages and accepted business records.
- [Human-in-the-loop agents: design an approval that means something](https://ontave.daltn.tech/index/human-in-the-loop-agents/index.md): Bind review to evidence, identity, a specific proposal and the effect a person is approving.
- [Structured LLM output is only the start of validation](https://ontave.daltn.tech/index/structured-output-validation/index.md): Check schema, source support, identity, permissions and current dependencies before committing facts.
- [Using LangGraph and Ontave in the same application](https://ontave.daltn.tech/index/langgraph-ontave/index.md): An architecture for execution checkpoints, durable evidence, explicit review and accepted-data reads.
- [Design an AI document-processing pipeline with reviewable records](https://ontave.daltn.tech/index/ai-document-processing/index.md): Preserve document revisions, handle retries, stage extracted claims and query accepted facts.
- [How to evaluate an AI agent beyond a successful demo](https://ontave.daltn.tech/index/evaluating-ai-agents/index.md): A concrete test plan for tools, task correctness, recovery, cost and human review effort.
- [A review boundary for agent workflows.](https://ontave.daltn.tech/index/agent-workflows/index.md): Follow a solar installation message from preserved evidence to accepted operational data.
- [Retrieval finds context. Ontave governs records.](https://ontave.daltn.tech/index/ontave-and-retrieval/index.md): Where retrieval, ordinary databases and an evidence-to-record framework each fit.
- [A changing model. A stable application contract.](https://ontave.daltn.tech/index/schema-evolution/index.md): Add an unfamiliar concept while existing applications keep their pinned interfaces.

## Other formats
- [Full index text](https://ontave.daltn.tech/llms-full.txt)
- [RSS feed](https://ontave.daltn.tech/feed.xml)
- [JSON feed](https://ontave.daltn.tech/feed.json)
- [Source and license](https://github.com/smileplis/ontave): Apache 2.0.