# Deyta

## Docs

- [API reference](https://docs.deyta.ai/khora/api-reference.md): The stable public surface of the Khora facade: construction, namespaces, remember/recall/forget, entity reads, result types, and the engine registry.
- [Overview](https://docs.deyta.ai/khora/architecture/overview.md): How Khora fits together: the facade, the three storage roles, the ingest and query pipelines, and the cross-cutting concerns.
- [Performance & scaling](https://docs.deyta.ai/khora/architecture/performance.md): The knobs that matter when ingesting and querying at scale: bulk loading, smart-mode resolution, connection pooling, batching, and HNSW tuning.
- [Rust acceleration](https://docs.deyta.ai/khora/architecture/rust-acceleration.md): An optional native extension (khora-accel) that speeds up Khora's CPU-bound operations, with automatic fallback to NumPy and pure Python.
- [Configuration](https://docs.deyta.ai/khora/configuration.md): Environment variables and KhoraConfig: every knob the library exposes.
- [Data model](https://docs.deyta.ai/khora/data-model.md): How data lives in Khora: documents, chunks, entities, relationships, episodes, and the immutable event log that ties them together.
- [Dream phase](https://docs.deyta.ai/khora/dream-phase.md): Offline maintenance pass for accumulated memory: entity dedupe, edge pruning, and graph hygiene, with a plan/apply split and undo snapshots.
- [Tuning](https://docs.deyta.ai/khora/engines/tuning.md): Tunable parameters for the VectorCypher engine: defaults, when to adjust them, and what you trade off.
- [VectorCypher](https://docs.deyta.ai/khora/engines/vectorcypher.md): Khora's hybrid retrieval engine: vector + Cypher graph + BM25 with query routing and RRF fusion.
- [Event sourcing](https://docs.deyta.ai/khora/event-sourcing.md): Every change in Khora is an immutable event in an append-only log, enabling audit trails, time travel, change-data-capture, and replay.
- [Basics](https://docs.deyta.ai/khora/examples/basics.md): Four runnable tutorials that teach the core loop: remember, recall, abstain, forget, and isolate by namespace.
- [Core APIs](https://docs.deyta.ai/khora/examples/core-apis.md): The five calls you reach for after the basics: batch ingest, recall filters, ontology control, and reading the graph back.
- [Workloads](https://docs.deyta.ai/khora/examples/workloads.md): VectorCypher end-to-end scenarios that compose Khora's APIs into real applications: a support-ticket knowledge graph, namespace versioning, and cross-document entity resolution.
- [CrewAI](https://docs.deyta.ai/khora/integrations/crewai.md): KhoraMemory: a drop-in storage backend for CrewAI's unified Memory.
- [Examples](https://docs.deyta.ai/khora/integrations/examples.md): Runnable, end-to-end demos of the first-party adapters (LangGraph, OpenAI Agents SDK, and CrewAI) that the reference pages describe.
- [Google ADK](https://docs.deyta.ai/khora/integrations/google-adk.md): KhoraMemoryService: BaseMemoryService drop-in for the ADK Runner.
- [Hermes](https://docs.deyta.ai/khora/integrations/hermes.md): KhoraMemoryProvider: Khora as the long-term memory plane for the Hermes agent framework.
- [LangGraph](https://docs.deyta.ai/khora/integrations/langgraph.md): KhoraStore: BaseStore implementation for StateGraph semantic long-term memory.
- [LlamaIndex](https://docs.deyta.ai/khora/integrations/llamaindex.md): Khora as a LlamaIndex retriever / vector store / memory.
- [OpenAI Agents SDK](https://docs.deyta.ai/khora/integrations/openai-agents.md): KhoraSession, khora_recall_tool, and KhoraMemoryHooks: session memory, recall-as-tool, and auto-persist.
- [Integrations](https://docs.deyta.ai/khora/integrations/overview.md): First-party adapters for CrewAI, LangGraph, Google ADK, OpenAI Agents, and LlamaIndex.
- [Khora](https://docs.deyta.ai/khora/introduction.md): Our open-source python library for creating knowledge repositories that ingest unstructured and structured multi-source data and expose a single query substrate, built for integrating into long-horizon AI agents.
- [Migrations](https://docs.deyta.ai/khora/migrations.md): How Khora's database schema is created and kept current: Alembic for PostgreSQL and the portable subset for sqlite_lance.
- [Namespaces & isolation](https://docs.deyta.ai/khora/namespaces.md): The namespace is Khora's only tenancy boundary: the dual-ID model, the query-layer isolation contract, versioning, and per-namespace config.
- [Observability](https://docs.deyta.ai/khora/observability.md): OpenTelemetry surface: [otel], [otel-grpc], and [logfire] extras, env-var contract, and exporter recipes.
- [Expertise & ontologies](https://docs.deyta.ai/khora/pipeline/expertise.md): An ExpertiseConfig is the domain ontology that drives extraction: typed entities and relationships, cross-source dedup, inferred edges, confidence, and prompts, bundled and reusable.
- [Semantic hooks](https://docs.deyta.ai/khora/pipeline/hooks.md): Subscribe to ingestion and recall events and get callbacks when the entities, relationships, or chunks they carry match your criteria.
- [Ingestion](https://docs.deyta.ai/khora/pipeline/ingestion.md): The write path: how remember() turns raw content into chunks, embeddings, and an entity graph through Khora's three-phase pipeline.
- [Retrieval](https://docs.deyta.ai/khora/pipeline/retrieval.md): The read path: how recall() understands a query, searches vector, graph, and keyword channels in parallel, and fuses the results.
- [Quickstart](https://docs.deyta.ai/khora/quickstart.md): Install Khora, bring up storage, and run your first remember / recall.
- [Storage backends](https://docs.deyta.ai/khora/storage-backends.md): Where Khora's data physically lives: the production PostgreSQL + pgvector + Neo4j stack and the embedded sqlite_lance alternative.

## OpenAPI Specs

- [openapi](https://docs.deyta.ai/api-reference/openapi.json)