Skip to content

chore(inference): xAI Grok causing problems, switching to OpenAI GPT Sol - #7913

Merged
rh-hemartin merged 1 commit into
mainfrom
change-to-openai
Oct 1, 2026
Merged

rh-hemartin merged 1 commit into
mainfrom
change-to-openai

Conversation

@rh-hemartin

Copy link
Copy Markdown
Member

Summary

xAI Grok was slow and barely working today, switching to OpenAI GPT Sol.

Related Issue

Changes

Testing

  • make lint passes (stage changes first, then run)
  • Tests added/updated for new or modified logic

Checklist

  • PR title follows Conventional Commits (correct type, ! for breaking changes)
  • Commits are signed off (DCO) — human and human-directed agent sessions only
  • I wrote this contribution myself and can explain all changes in it

@qodo-code-review

Copy link
Copy Markdown

PR Summary by Qodo

Switch code and review subagents from Grok to GPT Sol

⚙️ Configuration changes 🕐 Less than 10 minutes

Grey Divider

AI Description

• Switch the code agent to OpenAI GPT Sol after Grok reliability and latency issues.
• Switch the challenger, correctness, and security review subagents; leave other model assignments
 unchanged.
Diagram

graph TD
  Config["Fullsend config"] --> Code["Code agent"] --> Model["OpenAI GPT Sol"]
  Config --> Review["Review agent"] --> Subagents["Three review subagents"] --> Model
Loading
High-Level Assessment

The targeted configuration change is appropriate for an immediate provider switch. Automatic fallback would require broader orchestration changes without a demonstrated need in this PR.

Files changed (1) +4 / -4

Other (1) +4 / -4
config.yamlRoute four agent assignments to GPT Sol +4/-4

Route four agent assignments to GPT Sol

• Replaces xai/grok-4.6 with openai/gpt-6.1-sol for the code agent and the challenger, correctness, and security review subagents. Other agent and subagent model assignments remain unchanged.

.fullsend/config.yaml

@qodo-code-review

qodo-code-review Bot commented Oct 1, 2026 •

Copy link
Copy Markdown

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 🔗 Cross-repo conflicts (0) 📜 Skill insights (0)

Grey Divider


Action required

1. Reviews fail before subagents can start ✓ Resolved
Description
The review agent keeps a Sonnet parent while assigning openai/gpt-6.1-sol to three personas, but
pi’s trusted child-model set does not include OpenAI models unless they are the parent model. When a
review starts, persona resolution rejects those configured models; the runner also skips
provisioning the OpenAI provider because it checks only the parent model.
Code

.fullsend/config.yaml[R36-37]

+        challenger: openai/gpt-6.1-sol
+        correctness: openai/gpt-6.1-sol
Relevance

●●● Strong

Recent precedent accepts provider selection fixes for mixed parent and subagent models causing
runtime startup failures.

PR-#7718

ⓘ Recommendations generated based on similar findings in past PRs

Evidence
The changed review personas name an OpenAI model, but the unchanged parent is Sonnet. Bootstrap
builds its trusted set from the model table, registered provider models and parent model, then fails
resolution for a configured persona whose model is absent; the registered provider-model list
contains Google Vertex and xAI Vertex, not OpenAI. OpenAI provider creation is separately
conditional on the effective parent pi model.

.fullsend/config.yaml[33-41]
internal/runtime/pi_bootstrap.go[518-526]
internal/runtime/pi_bootstrap.go[657-668]
internal/runtime/pi_personas.go[230-238]
internal/runtime/pi_personas.go[288-295]
internal/runtime/pi_personas.go[393-416]
internal/runtime/openai_seeder.go[26-49]
internal/cli/run.go[1468-1487]

Agent prompt
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The Sonnet-parented review agent cannot start with OpenAI personas: pi rejects their models during bootstrap and the runner does not provision OpenAI credentials for child-only OpenAI usage.

## Fix Focus Areas
- .fullsend/config.yaml[33-41]
- internal/runtime/pi_bootstrap.go[657-668]
- internal/runtime/openai_seeder.go[26-49]
- internal/cli/run.go[1468-1487]

## Recommended Fix
Either use OpenAI as the review parent so the existing trusted-model and credential paths apply, or extend child-model admission and provider provisioning to account for configured OpenAI personas while retaining the Sonnet parent. Add a test that bootstraps and provisions a Sonnet-parented review with OpenAI children.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools


Grey Divider

Context sources
✅ Compliance rules (platform): 72 rules
✅ Cross-repo context — repo relationships
  Explored: repo: fullsend-ai/agents (sha: 3af27dde) — View relationship
Review mode: ⚖️ Balanced: This is a small but behavior-affecting configuration change that switches inference models, including correctness and security subagents, so it warrants a careful single-pass review.

Grey Divider

Tip of the day
💡 Did you know, you can keep summaries lean with Findings visible per group, which tucks the rest behind a View link

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗

Grey Divider

Qodo Logo

Comment thread .fullsend/config.yaml Resolved
@fullsend-ai-review

fullsend-ai-review Bot commented Oct 1, 2026 •

Copy link
Copy Markdown

🤖 Review · ⚠️ Cancelled · Started 2:33 PM UTC · Ended 2:39 PM UTC

Commit: e9a5290 · View workflow run →

@rh-hemartin
rh-hemartin enabled auto-merge October 1, 2026 14:35

@waynesun09 waynesun09 left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

code → openai/gpt-6.1-sol looks good. The review change needs one more line, though. As written, every review run fails at bootstrap.

With the review parent still on sonnet (Vertex), pi only accepts sub-agent models that are Claude aliases, Vertex Gemini, xai-vertex/xai/grok-4.6, or the parent's own model (piTrustedSpecs, internal/runtime/pi_personas.go). challenger/correctness/security are named in config, so the out-of-set model is fatal:

persona "security": resolved model "openai/gpt-6.1-sol" is not available in this run; accepted: anthropic-vertex/..., google-vertex/..., xai-vertex/xai/grok-4.6

Also, NeedsOpenAIProvider only looks at the parent model, so a Vertex parent would never get the OpenAI provider for those children anyway.

Fix: move the review parent to OpenAI too (suggestion inline). I checked this on main: OpenAI personas resolve, the Gemini personas still resolve, and NeedsOpenAIProvider is true.

Comment thread .fullsend/config.yaml Resolved
@fullsend-ai-review

fullsend-ai-review Bot commented Oct 1, 2026 •

Copy link
Copy Markdown

🤖 Review · ⚠️ Cancelled · Started 2:40 PM UTC · Ended 2:47 PM UTC

Commit: 2f3217c · View workflow run →

@codecov

codecov Bot commented Oct 1, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

Comment thread .fullsend/config.yaml Resolved
@fullsend-ai-review

fullsend-ai-review Bot commented Oct 1, 2026 •

Copy link
Copy Markdown

🤖 Review · ⚠️ Cancelled · Started 2:49 PM UTC · Ended 2:52 PM UTC

Commit: 6e6e232 · View workflow run →

Comment thread .fullsend/config.yaml Resolved
@fullsend-ai-review

fullsend-ai-review Bot commented Oct 1, 2026 •

Copy link
Copy Markdown

🤖 Review · ⚠️ Cancelled · Started 2:54 PM UTC · Ended 3:04 PM UTC

Commit: 2993e7f · View workflow run →

Signed-off-by: Hector Martinez <hemartin@redhat.com>
@waynesun09
waynesun09 dismissed their stale review October 1, 2026 15:04

Resolved: the review parent moved to openai/gpt-6.1-sol, so the OpenAI sub-agents are allowed and get the OpenAI provider.

@fullsend-ai-review

fullsend-ai-review Bot commented Oct 1, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 3:05 PM UTC · Completed 3:44 PM UTC

Commit: f33f8c0 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5 · Effort: high · Cost: $2.40

@rh-hemartin
rh-hemartin requested a review from waynesun09 October 1, 2026 15:09
@rh-hemartin
rh-hemartin added this pull request to the merge queue Oct 1, 2026
Merged via the queue into main with commit a017810 Oct 1, 2026
21 checks passed
@rh-hemartin
rh-hemartin deleted the change-to-openai branch October 1, 2026 15:10
@fullsend-ai-review

Copy link
Copy Markdown

Review skipped — this PR is already merged.

The /fs-review command only reviews open PRs/MRs.

Posted by fullsend post-review check

This branch was previously deployed

1 inactive deployment
dev — f33f8c06 Deployed Oct 1, 2026 by rh-hemartin via behaviour #14817
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants