Set scoped models for home Pi agent - #91
Conversation
Add enabledModels to pi_agent_settings_overrides for home site: - OpenAI: GPT-5.5 (openai-codex) - OpenCode: GLM-5, GLM-5.1, Kimi K2.5/K2.6, MiMo-V2.5/Pro, MiniMax M2.5/M2.7/M3, Qwen3.6 Plus/3.7 Plus/Max, DeepSeek V4 Pro/Flash (all via opencode-go provider)
📝 WalkthroughWalkthroughA single Ansible variable is added to the home host configuration, defining agent behavior settings including model allowlist, changelog display, and telemetry options. ChangesPI Agent Settings Configuration
Estimated code review effort🎯 1 (Trivial) | ⏱️ ~5 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
boga
changed the title
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@host_vars/home/vars.yml`:
- Around line 44-66: The pi_agent_settings_overrides block lists enabledModels
including openai-codex/gpt-5.5 which may be unavailable or return
model-not-found errors; update the code that consumes
pi_agent_settings_overrides (the enabledModels handling/validation logic) to
validate model IDs against an authoritative provider catalog (or a local
whitelist) before use, and add a runtime guard/fallback for openai-codex/gpt-5.5
specifically: detect failures and fall back to an alternative model or disable
that entry, and emit a clear warning/error when openai-codex/gpt-5.5 is selected
but not reachable; ensure the validation/fallback is invoked wherever
enabledModels is parsed so callers using enabledModels receive safe, validated
model selections.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 00996540-8b20-46d1-a3b6-c419fe3fdd89
📒 Files selected for processing (1)
host_vars/home/vars.yml
| pi_agent_settings_overrides: |- | ||
| { | ||
| "collapseChangelog": true, | ||
| "enableInstallTelemetry": false, | ||
| "enabledModels": [ | ||
| "openai-codex/gpt-5.5", | ||
| "opencode-go/glm-5", | ||
| "opencode-go/glm-5.1", | ||
| "opencode-go/kimi-k2.5", | ||
| "opencode-go/kimi-k2.6", | ||
| "opencode-go/mimo-v2.5", | ||
| "opencode-go/mimo-v2.5-pro", | ||
| "opencode-go/minimax-m2.5", | ||
| "opencode-go/minimax-m2.7", | ||
| "opencode-go/minimax-m3", | ||
| "opencode-go/qwen3.6-plus", | ||
| "opencode-go/qwen3.7-plus", | ||
| "opencode-go/qwen3.7-max", | ||
| "opencode-go/deepseek-v4-pro", | ||
| "opencode-go/deepseek-v4-flash" | ||
| ] | ||
| } | ||
|
|
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Search for model ID references, provider configuration, or documentation
# that confirms these models are valid
# Look for provider configuration files that might list available models
fd -t f -e json -e yaml -e yml 'models|providers|config' --exec rg -l 'openai-codex|opencode-go'
# Search for any documentation mentioning these specific models
rg -i 'gpt-5\.5|glm-5|kimi-k2|mimo-v2|minimax-m|qwen3\.[67]|deepseek-v4' --type md --type txt -C 2Repository: boga/dotfiles
Length of output: 39
🌐 Web query:
Are the following AI model IDs valid and currently available: openai-codex gpt-5.5, opencode-go glm-5.1, kimi-k2.6, mimo-v2.5-pro, minimax-m3, qwen3.7-max, deepseek-v4-pro?
💡 Result:
As of June 10, 2026, the status of the requested AI model IDs is as follows: openai-codex gpt-5.5: Valid but currently experiencing significant service availability issues [1][2]. While officially released for ChatGPT and Codex users [2], numerous reports indicate that users encounter 400 (unsupported model) and 404 (model not found) errors when attempting to use it via API or integrated tools, suggesting ongoing routing or entitlement propagation problems [3][4]. opencode-go glm-5.1: Valid and available [5][6]. It is a supported model within the OpenCode Go platform [5], although users may encounter "model not supported" errors if their account entitlements (subscription status) are not properly recognized or if they attempt to misconfigure the model identifier in their client settings [7][8]. kimi-k2.6: Valid and available [9]. It is accessible via the official Kimi platform, API, and various third-party providers [9][10]. It is also open-weight and available on Hugging Face [11]. mimo-v2.5-pro: Valid and available [12]. It was released by Xiaomi and is fully rolled out across their API platforms and accessible via Hugging Face [12]. minimax-m3: Valid and available [13][10]. It is supported by various providers and platforms, including OpenRouter and other API aggregators [13][10]. qwen3.7-max: Valid and available [14][15]. This model is offered by Qwen Cloud and multiple third-party API providers, often labeled as Qwen/Qwen3.7-Max or qwen-3.7-max [16][15][17]. deepseek-v4-pro: Valid and available [18]. This flagship model is provided via the official DeepSeek API and is available as open-weight model weights on Hugging Face [18].
Citations:
- 1: https://developers.openai.com/codex/models
- 2: https://openai.com/index/introducing-gpt-5-5/
- 3: gpt-5.5 is listed as available locally but real requests fail with 404 'Model not found' in both Desktop and CLI openai/codex#26892
- 4: gpt-5.5 shown in Codex, but rejected for ChatGPT Plus account with 400 unsupported-model error openai/codex#19654
- 5: https://dev.opencode.ai/docs/go/
- 6: https://whichllm.io/models/opencode-go-glm-5-1
- 7: 401 Model opencode-go/GLM-5.1 not supported anomalyco/opencode#23777
- 8: fix(models): add glm-5.1 to opencode-go catalogs NousResearch/hermes-agent#11269
- 9: https://www.kimi.com/ai-models/kimi-k2-6
- 10: https://aigateway.sh/compare/minimax_m3-vs-moonshot_kimi-k2.6
- 11: https://huggingface.co/moonshotai/Kimi-K2.6
- 12: https://mimo.xiaomi.com/mimo-v2-5-pro/
- 13: https://www.aimadetools.com/blog/minimax-m3-vs-kimi-k2-6/
- 14: https://www.qwencloud.com/models/qwen3.7-max-2026-05-20
- 15: https://www.together.ai/models/qwen37-max
- 16: https://llm24.net/model/qwen3-7-max
- 17: https://docs.kymaapi.com/models/qwen-3.7-max
- 18: https://deepseeksr1.com/v4-pro/
Confirm model IDs via external docs; add guardrails for openai-codex/gpt-5.5 availability
- No provider/model catalog for
openai-codex/opencode-gowas found in this repo, so validity relies on external references. - Public docs indicate the referenced models (e.g.,
glm-5.1,kimi-k2.6,mimo-v2-5-pro,minimax-m3,qwen3.7-max,deepseek-v4-pro) are real/available on their respective providers. openai-codex/gpt-5.5has documented ongoing API/model-not-found or entitlement/routing issues; ensure callers have fallback or error handling since this ID is included inenabledModels(and may be used by overrides).
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@host_vars/home/vars.yml` around lines 44 - 66, The
pi_agent_settings_overrides block lists enabledModels including
openai-codex/gpt-5.5 which may be unavailable or return model-not-found errors;
update the code that consumes pi_agent_settings_overrides (the enabledModels
handling/validation logic) to validate model IDs against an authoritative
provider catalog (or a local whitelist) before use, and add a runtime
guard/fallback for openai-codex/gpt-5.5 specifically: detect failures and fall
back to an alternative model or disable that entry, and emit a clear
warning/error when openai-codex/gpt-5.5 is selected but not reachable; ensure
the validation/fallback is invoked wherever enabledModels is parsed so callers
using enabledModels receive safe, validated model selections.
1 participant
Why
The home machine's Pi agent had no
enabledModelssetting, so the model pickershowed every built-in model. Scoping it to a specific list keeps the picker clean
and ensures only the intended providers — OpenAI Codex and OpenCode — are available
for interactive use and subagent dispatch.
Approach
Added
pi_agent_settings_overridestohost_vars/home/vars.ymlfollowing the samepattern already used on the work site. The OpenCode models are all served by the
built-in
opencode-goprovider (opencode.ai/zen/go/v1), so no custommodels.jsonentry is needed.
How it works
The Ansible
pi_agent/tasks/settings.ymltask deep-mergespi_agent_settings_overridesinto
~/.pi/agent/settings.jsonon the home host. The newenabledModelsarray restrictsthe picker to:
openai-codex/gpt-5.5opencode-go/glm-5,glm-5.1opencode-go/kimi-k2.5,kimi-k2.6opencode-go/mimo-v2.5,mimo-v2.5-proopencode-go/minimax-m2.5,minimax-m2.7,minimax-m3opencode-go/qwen3.6-plus,qwen3.7-plus,qwen3.7-maxopencode-go/deepseek-v4-pro,deepseek-v4-flashSummary by CodeRabbit
New Features
Chores