odysseus-dev/odysseus · error · Error

Server did not return a session id

Error message

Server did not return a session id

What it means

Raised by _fetch_bytes when a downloaded file's content length exceeds MAX_FILE_BYTES (400,000 bytes, ~400 KB). Every raw-asset fetch in the importer — SKILL.md itself, referenced files, and per-file downloads during directory traversal — goes through this single guard, so any one oversized file aborts the whole import.

Source

Thrown at src/visual_report.py:1096

  if (chatBtn) {{
    chatBtn.addEventListener('click', function() {{
      var researchId = chatBtn.dataset.researchId;
      if (!researchId) return;
      var origLabel = chatBtn.innerHTML;
      chatBtn.disabled = true;
      chatBtn.innerHTML = '<span>Creating chat…</span>';
      fetch('/api/research/spinoff/' + encodeURIComponent(researchId), {{
        method: 'POST', credentials: 'same-origin',
      }}).then(function(res) {{
        if (!res.ok) {{
          return res.json().then(function(d) {{
            throw new Error(d && d.detail ? d.detail : ('HTTP ' + res.status));
          }}, function() {{ throw new Error('HTTP ' + res.status); }});
        }}
        return res.json();
      }}).then(function(data) {{
        if (!data || !data.session_id) {{
          throw new Error('Server did not return a session id');
        }}
        var url = '/#' + data.session_id;
        var opened = false;
        // The report typically opens in a new tab — if we have access to the
        // original Odysseus tab, navigate it and close this report tab so the
        // user lands directly in the new chat.
        try {{
          if (window.opener && !window.opener.closed) {{
            window.opener.location.href = url;
            window.opener.location.reload();
            window.opener.focus();
            opened = true;
            window.close();
          }}
        }} catch (e) {{ /* cross-origin or detached opener — fall through */ }}
        if (!opened) {{
          // No opener (report was opened directly via URL) — open the chat in a
          // new tab so the report stays available.

View on GitHub (pinned to f9235ebbf1)

Solutions

1. Move the oversized asset out of the skill folder and reference it by URL inside SKILL.md instead of shipping it in the bundle
2. Split the large file into smaller pieces under 400 KB each, or trim it to what the skill actually needs
3. If you control the importer and genuinely need bigger files, raise MAX_FILE_BYTES in services/memory/skill_importer.py:21 — but note the 2 MB MAX_TOTAL_BYTES bundle cap still applies
4. Check for accidentally committed build artifacts (dist/, minified bundles, lock files) and remove them from the repo path

Example fix

# before
my-skill/
  SKILL.md
  big-dataset.json   # 1.2 MB -> SkillImportError: file too large

# after
my-skill/
  SKILL.md           # references https://example.com/data/big-dataset.json

import httpx
from services.memory.skill_importer import MAX_FILE_BYTES

async def fetch_if_small(url: str) -> bytes | None:
    async with httpx.AsyncClient(follow_redirects=True) as c:
        r = await c.head(url, follow_redirects=True)
        size = int(r.headers.get("content-length") or 0)
        if size and size > MAX_FILE_BYTES:
            return None  # skip before downloading
        r = await c.get(url)
        return r.content if len(r.content) <= MAX_FILE_BYTES else None

try:
    files, src = fetch_skill_bundle(url)
except SkillImportError as e:
    if "file too large" in str(e):
        report(f"{url}: an asset exceeds {MAX_FILE_BYTES} bytes — slim the skill folder or link assets by URL")
        return None
    raise

When it happens

Trigger: A skill folder contains a single text asset (large CSV dataset, bundled JSON, long reference doc, minified JS) larger than 400 KB; fetch_skill_bundle downloads its download_url via _fetch_bytes and len(r.content) > 400000. Directly fetching an oversized SKILL.md from a /blob/... link hits the same check.

Common situations: Skills that bundle prompt libraries, embeddings, dictionaries, or generated reference data inline instead of linking out. Also common after a doc-generation step adds a huge auto-produced file next to SKILL.md.

Related errors

• d.detail || 'Failed'
• d && d.detail ? d.detail : ('HTTP ' + res.status)
• {context} must stay on GitHub (got {host or 'unknown host'})
• skills.sh did not redirect to GitHub — open the skill's repo
• Invalid raw GitHub URL

AI-assisted analysis of odysseus-dev/odysseus@f9235ebbf1 (2026-08-14). Data as JSON: /api/errors/7810f0b32727ab91. Report an issue: GitHub.