HetaWiki
HetaWiki is Heta's LLM-native wiki layer. It compiles uploaded source documents into a versioned Markdown knowledge base and lets both users and agents browse, merge, and query that wiki directly.
Supported Formats
HetaWiki currently accepts the same lightweight document set used by its ingest pipeline:
| Format | Extensions |
|---|---|
| Plain text / markup | .md, .markdown, .txt |
.pdf |
|
| Word documents | .docx, .doc |
| Presentations | .pptx, .ppt |
LibreOffice is required to convert .doc, .ppt, and .pptx into PDF before
parsing.
Core Concepts
| Component | Role |
|---|---|
raw/ |
Immutable source files uploaded by the user |
wiki/pages/ |
Markdown pages that act as the source of truth for downstream reading and querying |
index.md |
Global wiki directory used by both the UI and the LLM |
log.md |
Append-only operation log for ingest, merge, and maintenance tasks |
| standalone git repo | Tracks wiki evolution independently from the main Heta codebase |
Two Ingest Modes
| Mode | What it does | Best for |
|---|---|---|
default ingest |
Parses one source document and turns it into exactly one new wiki page | Fast document onboarding |
merge ingest |
Runs an agent loop against a working copy of the wiki, reading and updating existing pages or creating new ones as needed | High-quality integration into an existing wiki |
Both modes are asynchronous. The API returns a task_id; callers poll the task
endpoint until the operation completes.
Query and Navigation
HetaWiki exposes three complementary ways to consume the wiki:
POST /api/v1/hetawiki/query— asynchronous wiki question answeringGET /api/v1/hetawiki/index— fetch the current wiki directoryGET /api/v1/hetawiki/pages/{filename}— read a single pageGET /api/v1/hetawiki/graph— materialise wiki links as nodes and edges for graph visualisation
This makes HetaWiki usable both as a Web UI feature and as an agent-readable knowledge layer.
Maintenance
HetaWiki also includes a configurable lint scheduler. Lint jobs scan the wiki periodically to surface structural issues such as inconsistent pages, missing links, or stale content candidates.
The scheduler is controlled via POST /api/v1/hetawiki/lint.
Related Pages
- Ingesting Documents — user guide for the frontend upload flow and ingest mode selection
- Querying the Wiki — ask questions, inspect source pages, and continue with follow-up questions
- REST API Reference
- Configuration Reference