Ingest anything.
Code, schemas, APIs, docs. Multimodal ingestion. everything your enterprise produces, parsed, chunked, embedded, and linked to your ontology. AI from step one, not bolted on later.
What We Ingest
Code at function granularity.
Not line-level chunks.
tree-sitter reads source code and breaks it into structural components (functions, classes, methods) rather than treating it as raw text. AST-based chunking guarantees every chunk is a complete, meaningful unit.
9 Languages · AST-Level Parsing
Plus: database schemas, API contracts (OpenAPI/Swagger, GraphQL), BPMN workflows, documentation (Confluence, Markdown). Every chunk gets a content-addressed SHA-256 ID.
Local AI First
AI from step one. Not bolted on later.
FastEmbed
BAAI/bge-small-en-v1.5. 384 dimensions. ONNX runtime. CPU-only. No cloud API calls.
GLiNER
Zero-shot NER. Technologies, organizations, standards. No training per entity type.
spaCy EntityRuler
Hand-written patterns for known entities. Deterministic matching complementing GLiNER.
Two Ingestion Modes
Fast mode or enriched mode.
You choose.
Fast Mode
Local math only. Steps 1–7. Parse, embed, link, propagate.
- AST parsing + chunking
- Local embedding + NER
- HyDE concept linking
- Hierarchy propagation
Enriched Mode
DefaultFull 8-step pipeline. 4-pass LLM enrichment.
- Everything in Fast Mode
- Contextual embedding
- Semantic analysis
- Cross-reference
- Quality scoring
- Smart model routing
Plug-and-Play Connectors
Your stack. No custom middleware.
Source Code
APIs
Databases
Infrastructure
Project Management
Data Pipelines
This site uses cookies
We use essential cookies for the site to function and analytics cookies (Google Analytics) to understand how you use it. Analytics cookies are only activated with your consent. We do not track you across other websites. Your data is stored in the EU and processed in accordance with GDPR. Read our Privacy Policy
