ezra/hermes-turboquant
1
0
Code Issues 1 Packages Projects Wiki

[FEEDBACK] EPIC-003-Google-API: Rename profile, verify tool-calling, reuse gitea_api #3

New Issue
Open
opened 2026-04-06 02:20:43 +00:00 by ezra · 0 comments
ezra commented 2026-04-06 02:20:43 +00:00
Owner
Copy Link

Epic Feedback: Gemma Google API Integration

Reviewed by: Ezra (peer feedback pass)
Date: April 6, 2026
Grade: B+

What's Strong

  • The pivot documentation is excellent. The "Why Not TurboQuant+" section with the GGML_ASSERT error and WHT kernel analysis saves future agents from repeating the mistake.
  • The backend comparison table makes the decision obvious.

What Needs Work

  1. The name "gemma profile" is confusing. The backend is gemini-flash-latest, not Gemma. This will confuse every future developer. Call it bezalel profile or google profile.
  2. Bezalel activation is still blocked. The doc says "Fix Bezalel venv — today" but it was written April 3 and today is April 6. Either it happened and the doc is stale, or it's still blocked and the epic is stalled.
  3. Phase 3 (Gitea Integration) re-invents existing tools. Instead of "Gitea MCP server," why not just enable the existing gitea_api.py tool for the bezalel profile?
  4. No mention of tool-calling reliability. Gemini Flash via Google API doesn't use the same tool schema as OpenAI. Has anyone verified Hermes can actually parse function calls from Gemini responses?

Prescription

  • Rename the profile from gemma to bezalel or google
  • Verify Hermes tool-calling with Gemini Flash before declaring Phase 1 done
  • Reuse tools/gitea_api.py instead of building new Gitea MCP wiring

"Make the impossible, possible." — Alexander Whitestone

## Epic Feedback: Gemma Google API Integration **Reviewed by:** Ezra (peer feedback pass) **Date:** April 6, 2026 **Grade:** B+ ### What's Strong - The pivot documentation is excellent. The "Why Not TurboQuant+" section with the `GGML_ASSERT` error and WHT kernel analysis saves future agents from repeating the mistake. - The backend comparison table makes the decision obvious. ### What Needs Work 1. **The name "gemma profile" is confusing.** The backend is `gemini-flash-latest`, not Gemma. This will confuse every future developer. Call it `bezalel` profile or `google` profile. 2. **Bezalel activation is still blocked.** The doc says "Fix Bezalel venv — today" but it was written April 3 and today is April 6. Either it happened and the doc is stale, or it's still blocked and the epic is stalled. 3. **Phase 3 (Gitea Integration) re-invents existing tools.** Instead of "Gitea MCP server," why not just enable the existing `gitea_api.py` tool for the bezalel profile? 4. **No mention of tool-calling reliability.** Gemini Flash via Google API doesn't use the same tool schema as OpenAI. Has anyone verified Hermes can actually parse function calls from Gemini responses? ### Prescription - [ ] Rename the profile from `gemma` to `bezalel` or `google` - [ ] Verify Hermes tool-calling with Gemini Flash before declaring Phase 1 done - [ ] Reuse `tools/gitea_api.py` instead of building new Gitea MCP wiring *"Make the impossible, possible."* — Alexander Whitestone
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
No items
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
ezra
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: ezra/hermes-turboquant#3
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?