Document Formats

Council can ingest documents to train persona experts and enrich panel context. Support varies from native (text-based) to rich (converted) to AI-assisted (experimental).

Supported Formats

View the complete, up-to-date list from your current installation:

council docs formats

Native (Text-Based)

These formats are parsed directly with no conversion overhead.

Extension	Description
`.md`, `.markdown`	Markdown
`.txt`	Plain text
`.html`, `.htm`	HTML

Rich Documents (Converted to Text)

These formats are converted to plain text using specialized extractors.

Extension	Description	Notes
`.pdf`	PDF	Text extraction via native parser
`.docx`	Word document	OpenXML format
`.pptx`	PowerPoint presentation	Slide text + speaker notes
`.xlsx`	Excel spreadsheet	Sheet names, cell values
`.xls`	Legacy Excel	Re-save as `.xlsx` recommended
`.csv`	Comma-separated values	Column headers + rows
`.tsv`	Tab-separated values	Column headers + rows
`.rtf`	Rich Text Format	Text + basic formatting
`.odt`	OpenDocument Text	LibreOffice/OpenOffice
`.ods`	OpenDocument Spreadsheet	LibreOffice/OpenOffice
`.odp`	OpenDocument Presentation	LibreOffice/OpenOffice

File Size Limits

Default: 50 MB per file

Files exceeding this limit are rejected with an oversize-file error before any processing.

Configure:

council config set documents.maxFileSizeMB 100

Range: 1–500 MB

AI Extraction (Experimental)

For formats without a native extractor, Council can optionally use AI-based extraction to build a structured text description. This never sends files to an external service — AI extraction runs locally via the configured engine (GitHub Copilot).

Modes

Mode	Behavior
`off`	(Default) Reject unknown formats. No AI extraction.
`ask`	Prompt user for approval before extracting each file.
`auto`	Automatically extract unknown formats without prompting.

Configure:

council config set documents.aiExtraction ask

When to Enable:

Panels that need otherwise-unreadable files (e.g., proprietary formats, screenshots with text)
You want structured descriptions of images, diagrams, or scanned PDFs (OCR)
Experimenting with new document types

Performance Note: AI extraction is slower and consumes API tokens. Use sparingly for production workflows.

Whitelist Extensions

By default, AI extraction applies to all unknown formats. Restrict to specific extensions:

council config set documents.aiExtractionAllowedExtensions '[".png", ".jpg", ".bmp"]'

Empty array = all extensions eligible.

Checking Supported Formats

CLI Command

council docs formats

Output includes:

Native formats
Rich document formats
AI extraction status and mode
File size limit
Configuration instructions

Programmatic Check

import { getSupportedExtensions } from "@council-ai/cli/core/documents/extractors";

const extensions = getSupportedExtensions();
console.log(extensions); // [".md", ".txt", ".pdf", ...]

Corpus Management

After adding documents to a panel’s docs/ folder:

Review Pending Files

council docs review <panel>

Lists files that couldn’t be processed (unsupported format, extraction failed, AI-eligible).

Exit code: Non-zero if any files are pending review (useful in CI).

Health Check

council docs doctor <panel>

Shows:

Total indexed documents + word count
Pending review count
Corrupt file count
Configured AI extraction mode
File size limit

Re-Index Documents

council docs extract <panel>

Re-runs extraction on pending files (e.g., after enabling AI extraction or adding new documents).

Configuration Reference

Config Key	Type	Default	Description
`expert.supportedFormats`	string[]	(see above)	File extensions eligible for ingestion.
`documents.aiExtraction`	`off` \| `ask` \| `auto`	`off`	AI-based extraction fallback mode.
`documents.aiExtractionAllowedExtensions`	string[]	`[]`	Whitelist for AI extraction. Empty = all.
`documents.maxFileSizeMB`	integer	50	Maximum file size (1–500 MB).

Configuration — Full config schema
Expert YAML Format — docsPath field for persona experts