Bug Description
PR #24013 (feat(ui-tui): resolve markdown links to readable page titles) introduced ResolvedLink + useLinkTitle() to replace bare URLs with fetched page titles in the TUI.
When the async title fetch fails or returns no result, the display falls back to urlSlugTitleLabel() — a function that parses the URL path and title-cases the last meaningful segment. This produces unintelligible single-word labels for many real-world URLs:
https://x.com/OpenAI/status/12345 → "Status"https://accounts.google.com/o/oauth2/auth?client_id=... → "Auth" or similar generic slug- Any URL where the path ends in a common generic word produces a label that hides the actual destination
The same session content in the Web UI correctly shows the full clickable hyperlink.
Steps to Reproduce
- Start a TUI session (
hermes in terminal).
- Ask the agent to summarize an X/Twitter timeline or include any tweet links in its response.
- Observe that every
https://x.com/.../status/... link is rendered as: 🔗 Status
- Open the same session in Web UI — the full URL is visible and clickable.
Expected Behavior
When title resolution fails or produces a generic slug, the TUI should fall back to the original URL string (e.g. https://x.com/OpenAI/status/1234567890) rather than a single ambiguous word.
Actual Behavior
The link is displayed as the literal text "Status". The numeric tweet ID is filtered out by the slug parser (\d+ regex in urlSlugTitleLabel), leaving only the preceding path segment.
Affected Component
Other, CLI (interactive chat)
Messaging Platform (if gateway-related)
No response
Debug Report
Not a runtime crash — this is a rendering / UX regression in URL display logic.
Operating System
Ubuntu 24.04.4 LTS
Python Version
3.11.15
Hermes Version
0.13.0
Additional Logs / Traceback (optional)
Root Cause Analysis (optional)
No response
Proposed Fix (optional)
- Blacklist common generic path segments in
urlSlugTitleLabel() — e.g. status, page, id, item, post, home. When the resolved slug matches one of these, fall back to the raw URL instead of the slug.
- Prefer raw URL over short/generic slug when the slug is ≤ 3 characters or is a known generic term.
- Optional: Expose a config toggle (
tui.resolve_link_titles) to disable title fetching entirely for users who prefer raw URLs.
Are you willing to submit a PR for this?
Bug Description
PR #24013 (
feat(ui-tui): resolve markdown links to readable page titles) introducedResolvedLink+useLinkTitle()to replace bare URLs with fetched page titles in the TUI.When the async title fetch fails or returns no result, the display falls back to
urlSlugTitleLabel()— a function that parses the URL path and title-cases the last meaningful segment. This produces unintelligible single-word labels for many real-world URLs:https://x.com/OpenAI/status/12345→ "Status"https://accounts.google.com/o/oauth2/auth?client_id=...→ "Auth" or similar generic slugThe same session content in the Web UI correctly shows the full clickable hyperlink.
Steps to Reproduce
hermesin terminal).https://x.com/.../status/...link is rendered as: 🔗 StatusExpected Behavior
When title resolution fails or produces a generic slug, the TUI should fall back to the original URL string (e.g.
https://x.com/OpenAI/status/1234567890) rather than a single ambiguous word.Actual Behavior
The link is displayed as the literal text "Status". The numeric tweet ID is filtered out by the slug parser (
\d+regex inurlSlugTitleLabel), leaving only the preceding path segment.Affected Component
Other, CLI (interactive chat)
Messaging Platform (if gateway-related)
No response
Debug Report
Not a runtime crash — this is a rendering / UX regression in URL display logic.Operating System
Ubuntu 24.04.4 LTS
Python Version
3.11.15
Hermes Version
0.13.0
Additional Logs / Traceback (optional)
Root Cause Analysis (optional)
No response
Proposed Fix (optional)
urlSlugTitleLabel()— e.g.status,page,id,item,post,home. When the resolved slug matches one of these, fall back to the raw URL instead of the slug.tui.resolve_link_titles) to disable title fetching entirely for users who prefer raw URLs.Are you willing to submit a PR for this?