odysseus-dev/odysseus · error · HTTPException

No vision model available: {e}

Error message

No vision model available: {e}

What it means

Error "No vision model available: {e}" thrown in odysseus-dev/odysseus.

Source

Thrown at routes/document/document_routes.py:1281

            if not doc:
                raise HTTPException(404, "Document not found")
            _verify_doc_owner(db, doc, user)
            upload_id = find_source_upload_id(doc.current_content or "")
            if not upload_id:
                raise HTTPException(400, "Document is not linked to a source PDF")
            pdf_path = _locate_current_user_upload(request, upload_id, user)
            if not pdf_path:
                raise HTTPException(404, "Source PDF not found")
        finally:
            db.close()

        # Resolve VL model (admin-configured or auto-detected vision-capable)
        settings = _load_vl_settings()
        vl_model = settings.get("vision_model", "")
        try:
            url, model_id, headers = _resolve_vl_model(vl_model, owner=user)
        except Exception as e:
            raise HTTPException(503, f"No vision model available: {e}")

        system_prompt = (
            "You analyze rendered PDF page images and propose values to fill in. "
            "For each blank line, box, underscore, or labeled space on the page that "
            "should be filled given the user's instruction, output one annotation. "
            "Coordinates are percentages (0-100) of the page width/height with the "
            "origin at top-left. Width/height should match the visible blank box. "
            "Return ONLY a JSON array, no prose, no markdown fences. Each entry: "
            '{"x": number, "y": number, "w": number, "h": number, "value": string}. '
            "If a region should not be filled, omit it. If nothing should be filled, "
            "return []."
        )

        all_annotations = []
        pdf_doc = fitz.open(pdf_path)
        try:
            for page_index in range(pdf_doc.page_count):
                page = pdf_doc[page_index]

View on GitHub (pinned to f9235ebbf1)

Solutions

1. Configure a vision-capable model endpoint in Settings before using this feature.
2. Check that the vision model service is running and reachable, then retry.

When it happens

Trigger: Triggered when the corresponding server-side validation or runtime check at the recorded location rejects the request or operation and returns this error message to the caller.

Common situations: See trigger scenarios.

AI-assisted analysis of odysseus-dev/odysseus@f9235ebbf1 (2026-08-14). Data as JSON: /api/errors/b3807bbe5b543ce2. Report an issue: GitHub.