Repository navigation
docs: add GMI Cloud as supported LLM provider (HackwithBay co-sponsor) - #489
josephmccann wants to merge 1 commit into
Conversation
Relates to #414 (co-sponsor integrations) GMI Cloud is a hackathon co-sponsor offering 40+ models via OpenAI-compatible API. Works with existing llm_openai node via base_url_env override. Documents: - Configuration for llm_openai node - Available models with use-case guidance - Multi-model parallel evaluation pattern
📝 WalkthroughWalkthroughA new documentation page for GMI Cloud provider was added, describing integration via the Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
There was a problem hiding this comment.
Actionable comments posted: 5
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@docs/providers/gmi-cloud.md`:
- Around line 44-47: The example uses unsupported node config keys (base_url_env
and api_key_env) in the llm_openai nodes (ids: deepseek, qwen), so update those
node entries to use the llm_openai schema's supported fields (for example set
explicit base_url and api_key entries or remove those fields and reference
environment variables externally), and ensure the multi-node flow feeding into
the compare node remains unchanged; locate the nodes array and replace
base_url_env/api_key_env with the correct schema keys or omit them and add a
comment showing how to provide credentials via environment/config for the
llm_openai nodes.
- Around line 24-29: The markdown code block showing environment variables
(GMI_API_KEY and GMI_BASE) lacks a language specifier; update the
triple-backtick fence for that block to include "bash" (i.e., ```bash) so the
snippet for GMI_API_KEY and GMI_BASE gets proper shell syntax highlighting in
docs/providers/gmi-cloud.md.
- Around line 61-66: Update the broken models catalog link in the "Getting
Started" section: replace the incorrect URL "https://console.gmicloud.ai/models"
referenced under the "Getting Started" header with the correct GMI Cloud models
path (confirm with GMI Cloud) or point to an alternative discovery method (e.g.,
console root or API docs). Ensure the doc still shows steps 1–4 and retains the
environment variable references GMI_API_KEY and GMI_BASE so users can find model
IDs via the validated link or the suggested alternative.
- Around line 8-22: The docs show unsupported fields (base_url_env, api_key_env)
for the llm_openai node; either update the docs to use the node's actual schema
(use "model" and provide the API key via llm.cloud.apikey / apikey and remove
base_url_env) or update the llm_openai implementation to accept and honor
base_url_env/api_key_env: add schema entries for base_url_env and api_key_env
(or a single baseUrl/base_url_env and apikey_env), read those env vars in the
node setup, and pass the resolved api key and base URL into the ChatOpenAI
initializer (extend the ChatOpenAI construction code in the llm_openai handler
to accept a baseUrl option and use llm.cloud.apikey if present). Ensure
references to modelTotalTokens and apikey (llm.cloud.apikey) remain compatible.
- Around line 33-39: The table rows containing incorrect model identifiers
(`deepseek-ai/DeepSeek-V3.2`, `Qwen/Qwen3.5-35B-A3B`, `MiniMaxAI/MiniMax-M2.7`,
`google/gemini-3.1-flash-lite-preview`, `anthropic/claude-sonnet-4-6`) must be
replaced with actual GMI Cloud model IDs; fetch the authoritative list via GET
https://api.gmi-serving.com/v1/models (or the GMI Cloud model listings) and
update the markdown table entries to the exact returned IDs (e.g.,
`deepseek-ai/DeepSeek-V3` or `deepseek-ai/DeepSeek-V3.1` instead of
`deepseek-ai/DeepSeek-V3.2`, and correct `Qwen/...` to the validated `Qwen`
variant), ensuring each table cell contains a valid model identifier that
matches the API response.
🪄 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: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 3b81f2e3-79be-4c51-b3d1-fe0f2b71ad43
📒 Files selected for processing (1)
docs/providers/gmi-cloud.md
| GMI Cloud works with the `llm_openai` node using a custom base URL: | ||
|
|
||
| ```json | ||
| { | ||
| "id": "analysis_node", | ||
| "type": "llm_openai", | ||
| "label": "Pattern Analysis (GMI Cloud)", | ||
| "config": { | ||
| "model": "deepseek-ai/DeepSeek-V3.2", | ||
| "base_url_env": "GMI_BASE", | ||
| "api_key_env": "GMI_API_KEY", | ||
| "system": "Your system prompt here" | ||
| } | ||
| } | ||
| ``` |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Verify the actual llm_openai node schema and implementation
echo "=== Checking node schema fields ==="
rg -A 50 '"fields"' nodes/src/nodes/llm_openai/services.json | head -80
echo -e "\n=== Checking ChatOpenAI initialization ==="
rg -B 5 -A 10 'ChatOpenAI\(' nodes/src/nodes/llm_openai/openai_client.py
echo -e "\n=== Checking config extraction ==="
rg -B 5 -A 10 'getNodeConfig' nodes/src/nodes/llm_openai/Repository: rocketride-org/rocketride-server
Length of output: 4756
Critical: Documented configuration fields do not exist in llm_openai node schema.
The configuration example uses base_url_env and api_key_env fields, but the llm_openai node implementation only supports:
model(string)modelTotalTokens(number)apikey(viallm.cloud.apikey)
The fields base_url_env and api_key_env are not defined in the node schema and will be silently ignored at runtime. Additionally, the ChatOpenAI initialization does not accept a base URL parameter, so there is no mechanism to override the base URL through the documented configuration.
This means the documented configuration will not work for GMI Cloud integration.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@docs/providers/gmi-cloud.md` around lines 8 - 22, The docs show unsupported
fields (base_url_env, api_key_env) for the llm_openai node; either update the
docs to use the node's actual schema (use "model" and provide the API key via
llm.cloud.apikey / apikey and remove base_url_env) or update the llm_openai
implementation to accept and honor base_url_env/api_key_env: add schema entries
for base_url_env and api_key_env (or a single baseUrl/base_url_env and
apikey_env), read those env vars in the node setup, and pass the resolved api
key and base URL into the ChatOpenAI initializer (extend the ChatOpenAI
construction code in the llm_openai handler to accept a baseUrl option and use
llm.cloud.apikey if present). Ensure references to modelTotalTokens and apikey
(llm.cloud.apikey) remain compatible.
| Set environment variables: | ||
|
|
||
| ``` | ||
| GMI_API_KEY=your_api_key_here | ||
| GMI_BASE=https://api.gmi-serving.com/v1/chat/completions | ||
| ``` |
There was a problem hiding this comment.
Minor: Add language specification to code block.
The bash/shell code block should specify a language for proper syntax highlighting.
📝 Proposed fix
Set environment variables:
-```
+```bash
GMI_API_KEY=your_api_key_here
GMI_BASE=https://api.gmi-serving.com/v1/chat/completions
</details>
**Note:** This environment variable configuration depends on resolving the critical issue with the `base_url_env` and `api_key_env` fields (lines 8-22).
<!-- suggestion_start -->
<details>
<summary>📝 Committable suggestion</summary>
> ‼️ **IMPORTANT**
> Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
```suggestion
Set environment variables:
🧰 Tools
🪛 markdownlint-cli2 (0.22.0)
[warning] 26-26: Fenced code blocks should have a language specified
(MD040, fenced-code-language)
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@docs/providers/gmi-cloud.md` around lines 24 - 29, The markdown code block
showing environment variables (GMI_API_KEY and GMI_BASE) lacks a language
specifier; update the triple-backtick fence for that block to include "bash"
(i.e., ```bash) so the snippet for GMI_API_KEY and GMI_BASE gets proper shell
syntax highlighting in docs/providers/gmi-cloud.md.
| | Model | Best For | | ||
| |---|---| | ||
| | `deepseek-ai/DeepSeek-V3.2` | Structured extraction, JSON output | | ||
| | `Qwen/Qwen3.5-35B-A3B` | Reasoning, analysis | | ||
| | `MiniMaxAI/MiniMax-M2.7` | Long context | | ||
| | `google/gemini-3.1-flash-lite-preview` | Fast, cheap | | ||
| | `anthropic/claude-sonnet-4-6` | High quality | |
There was a problem hiding this comment.
🧩 Analysis chain
🌐 Web query:
What are the correct model identifiers for GMI Cloud's API? Specifically verify: deepseek-ai/DeepSeek-V3.2, Qwen/Qwen3.5-35B-A3B, MiniMaxAI/MiniMax-M2.7, google/gemini-3.1-flash-lite-preview, anthropic/claude-sonnet-4-6
💡 Result:
GMI Cloud’s Inference Engine uses OpenAI-style model strings and recommends you list the exact available IDs via GET https://api.gmi-serving.com/v1/models, then pass that value as model in requests. [1]
Verification of the specific identifiers you provided (against GMI’s published model-ID lists):
| Provided identifier | Verified as a GMI Cloud model ID? | What GMI docs show instead |
|---|---|---|
deepseek-ai/DeepSeek-V3.2 |
Not found in GMI’s published model-ID lists | GMI lists IDs like deepseek-ai/DeepSeek-V3, deepseek-ai/DeepSeek-V3-0324, deepseek-ai/DeepSeek-V3.1, deepseek-ai/DeepSeek-V3.1-Terminus (and others). [2][3] |
Qwen/Qwen3.5-35B-A3B |
Not found in GMI’s published model-ID lists | GMI lists (examples) Qwen/Qwen3-30B-A3B, Qwen/Qwen3-32B-FP8, Qwen/Qwen3-Next-80B-A3B-Instruct, etc. [2][3] |
MiniMaxAI/MiniMax-M2.7 |
Not found in GMI’s published model-ID lists | No MiniMaxAI/... IDs appear in the referenced GMI model-ID lists. [2][3] |
google/gemini-3.1-flash-lite-preview |
Not found in GMI’s published model-ID lists | No google/... (Gemini) IDs appear in the referenced GMI model-ID lists. [2][3] |
anthropic/claude-sonnet-4-6 |
Not found in GMI’s published model-ID lists | Anthropic’s own Claude API model ID is claude-sonnet-4-6 (note: no anthropic/ prefix), but GMI does not list it in the referenced GMI model-ID lists. [4][2][3] |
What to do to get the “correct” IDs for your GMI account: call GET https://api.gmi-serving.com/v1/models and use the returned id values. [1]
Sources:
[1] GMI Cloud Docs — LLM API Reference (/v1/models, model field usage) (docs.gmicloud.ai)
[2] GMI Cloud Docs — Rate limits page (published model ID strings) (docs.gmicloud.ai)
[3] GMI Cloud Docs — Dify tutorial (example/preset model IDs) (docs.gmicloud.ai)
[4] Anthropic Claude API Docs — Models overview (Claude API IDs incl. claude-sonnet-4-6) (platform.claude.com)
Citations:
- 1: https://docs.gmicloud.ai/inference-engine/api-reference/llm-api-reference
- 2: https://docs.gmicloud.ai/inference-engine/api-reference/rate-limit
- 3: https://docs.gmicloud.ai/inference-engine/tutorials/build-deep-research-workflow-in-dify?utm_source=openai
- 4: https://platform.claude.com/docs/en/about-claude/models/overview?utm_source=openai
All five model identifiers in this table are incorrect and will cause runtime errors.
Verification against GMI Cloud's published documentation shows none of these IDs exist:
deepseek-ai/DeepSeek-V3.2→ should be variants likedeepseek-ai/DeepSeek-V3ordeepseek-ai/DeepSeek-V3.1Qwen/Qwen3.5-35B-A3B→ should beQwen/Qwen3-30B-A3Bor similarMiniMaxAI/MiniMax-M2.7→ not available in GMI Cloudgoogle/gemini-3.1-flash-lite-preview→ not available in GMI Cloudanthropic/claude-sonnet-4-6→ not available in GMI Cloud
Replace this table with correct IDs by calling GET https://api.gmi-serving.com/v1/models against the GMI Cloud API, or reference the official GMI Cloud model listings.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@docs/providers/gmi-cloud.md` around lines 33 - 39, The table rows containing
incorrect model identifiers (`deepseek-ai/DeepSeek-V3.2`,
`Qwen/Qwen3.5-35B-A3B`, `MiniMaxAI/MiniMax-M2.7`,
`google/gemini-3.1-flash-lite-preview`, `anthropic/claude-sonnet-4-6`) must be
replaced with actual GMI Cloud model IDs; fetch the authoritative list via GET
https://api.gmi-serving.com/v1/models (or the GMI Cloud model listings) and
update the markdown table entries to the exact returned IDs (e.g.,
`deepseek-ai/DeepSeek-V3` or `deepseek-ai/DeepSeek-V3.1` instead of
`deepseek-ai/DeepSeek-V3.2`, and correct `Qwen/...` to the validated `Qwen`
variant), ensuring each table cell contains a valid model identifier that
matches the API response.
| { | ||
| "nodes": [ | ||
| {"id": "deepseek", "type": "llm_openai", "config": {"model": "deepseek-ai/DeepSeek-V3.2", "base_url_env": "GMI_BASE", "api_key_env": "GMI_API_KEY"}}, | ||
| {"id": "qwen", "type": "llm_openai", "config": {"model": "Qwen/Qwen3.5-35B-A3B", "base_url_env": "GMI_BASE", "api_key_env": "GMI_API_KEY"}}, |
There was a problem hiding this comment.
Critical: Pipeline example uses non-existent configuration fields.
This multi-model evaluation example suffers from the same critical issue identified in lines 8-22: the base_url_env and api_key_env fields do not exist in the llm_openai node schema.
The parallel execution pattern (multiple nodes feeding into compare) is architecturally sound, but the node configurations will not function as written.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@docs/providers/gmi-cloud.md` around lines 44 - 47, The example uses
unsupported node config keys (base_url_env and api_key_env) in the llm_openai
nodes (ids: deepseek, qwen), so update those node entries to use the llm_openai
schema's supported fields (for example set explicit base_url and api_key entries
or remove those fields and reference environment variables externally), and
ensure the multi-node flow feeding into the compare node remains unchanged;
locate the nodes array and replace base_url_env/api_key_env with the correct
schema keys or omit them and add a comment showing how to provide credentials
via environment/config for the llm_openai nodes.
| ## Getting Started | ||
|
|
||
| 1. Sign up at [console.gmicloud.ai](https://console.gmicloud.ai) | ||
| 2. Generate an API key | ||
| 3. Set `GMI_API_KEY` and `GMI_BASE` in your environment | ||
| 4. Use any model ID from the [GMI model catalog](https://console.gmicloud.ai/models) |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Verify GMI Cloud URLs are accessible
echo "=== Checking GMI Cloud console ==="
curl -s -o /dev/null -w "HTTP %{http_code}\n" https://console.gmicloud.ai
echo "=== Checking GMI Cloud models page ==="
curl -s -o /dev/null -w "HTTP %{http_code}\n" https://console.gmicloud.ai/models
echo "=== Checking GMI Cloud API endpoint ==="
curl -s -o /dev/null -w "HTTP %{http_code}\n" https://api.gmi-serving.com/v1/chat/completionsRepository: rocketride-org/rocketride-server
Length of output: 216
Fix broken GMI Cloud models catalog URL in getting-started instructions.
The referenced models catalog URL at console.gmicloud.ai/models returns HTTP 404. Verify the correct path with GMI Cloud or update the documentation link. The console base URL (console.gmicloud.ai) is accessible, so ensure the models page URL is accurate or redirect users to an alternative method for discovering available model IDs.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@docs/providers/gmi-cloud.md` around lines 61 - 66, Update the broken models
catalog link in the "Getting Started" section: replace the incorrect URL
"https://console.gmicloud.ai/models" referenced under the "Getting Started"
header with the correct GMI Cloud models path (confirm with GMI Cloud) or point
to an alternative discovery method (e.g., console root or API docs). Ensure the
doc still shows steps 1–4 and retains the environment variable references
GMI_API_KEY and GMI_BASE so users can find model IDs via the validated link or
the suggested alternative.
|
Hey again, @josephmccann, as I mentioned in another PR, I know that it is trivial, but it will be good if the code example complies with the coderabbit advices |
Senior Review: docs — GMI Cloud as supported LLM providerWhat works well
Should fix
Nice-to-have
Straightforward doc addition — the main risk is inaccurate config field names leading to user confusion. Please verify against the actual schema. |
🚀 Merge RequestUseful provider documentation. @josephmccann — dsapandora requested CodeRabbit feedback be addressed:
Small fixes, low risk. Ready to merge after updates. |
|
Changes requested — ## Review: docs/providers/gmi-cloud.md Honest take: this PR has good intentions but lands in a rough spot. Documenting a sponsor integration for hackathon participants is a legitimate goal, but there are enough concrete problems here that I can't approve it as-is. The Big Problem: Superseded by PR #540The most glaring issue is that a dedicated The
|
Top-level `permissions: { contents: write, packages: write, id-token: write }`
on nightly.yaml granted every job in the workflow the union of all writes
any one job needed, including jobs that only consume read-only inputs. The
Scorecard `TokenPermissionsID` rule flags this as the most severe form of
over-grant because top-level writes propagate to every reuse-called
workflow as the upper bound.
Drop the top-level to `contents: read` and raise per-job only where the
job actually writes:
- build: contents: read, packages: write
(forwarded to _build.yaml for NuGet/vcpkg writes
to ghcr.io via VCPKG_BINARY_SOURCES=...readwrite)
- cleanup-prereleases: contents: write
(git push origin --delete <tag> and gh release
delete; the tag deletion requires repo write)
- prerelease: contents: write, id-token: write
(forwarded to _release.yaml for tag push +
gh release create + Sigstore keyless signing)
- docker: contents: read, packages: write
(forwarded to _docker.yaml for ghcr image push;
matches release.yaml's docker job pattern)
- init: no override, inherits top-level read
(_init.yaml is read-only — extracts versions
from package.json via jq and stamps github.sha)
The grants mirror the canonical pattern already in place in release.yaml
(committed in the prior TokenPermissionsID cleanup pass that closed #513
and #515 on _build.yaml and _init.yaml).
Closes Scorecard TokenPermissionsID alerts #566 (topLevel contents:write)
and #567 (topLevel packages:write).
The remaining 10 TokenPermissionsID alerts on this repo (#451, #452, #454,
#455, #489, #490, #604, #635, #649, #650) flag job-level writes that are
required for the job's function (ghcr image push, GitHub release publish,
artifact cleanup, model-sync commits). Those are being dismissed
separately with "won't fix" + per-alert justification — Scorecard
penalizes any write, including the unavoidable kind.
…-org#929) Top-level `permissions: { contents: write, packages: write, id-token: write }` on nightly.yaml granted every job in the workflow the union of all writes any one job needed, including jobs that only consume read-only inputs. The Scorecard `TokenPermissionsID` rule flags this as the most severe form of over-grant because top-level writes propagate to every reuse-called workflow as the upper bound. Drop the top-level to `contents: read` and raise per-job only where the job actually writes: - build: contents: read, packages: write (forwarded to _build.yaml for NuGet/vcpkg writes to ghcr.io via VCPKG_BINARY_SOURCES=...readwrite) - cleanup-prereleases: contents: write (git push origin --delete <tag> and gh release delete; the tag deletion requires repo write) - prerelease: contents: write, id-token: write (forwarded to _release.yaml for tag push + gh release create + Sigstore keyless signing) - docker: contents: read, packages: write (forwarded to _docker.yaml for ghcr image push; matches release.yaml's docker job pattern) - init: no override, inherits top-level read (_init.yaml is read-only — extracts versions from package.json via jq and stamps github.sha) The grants mirror the canonical pattern already in place in release.yaml (committed in the prior TokenPermissionsID cleanup pass that closed rocketride-org#513 and rocketride-org#515 on _build.yaml and _init.yaml). Closes Scorecard TokenPermissionsID alerts rocketride-org#566 (topLevel contents:write) and rocketride-org#567 (topLevel packages:write). The remaining 10 TokenPermissionsID alerts on this repo (rocketride-org#451, rocketride-org#452, rocketride-org#454, rocketride-org#455, rocketride-org#489, rocketride-org#490, rocketride-org#604, rocketride-org#635, rocketride-org#649, rocketride-org#650) flag job-level writes that are required for the job's function (ghcr image push, GitHub release publish, artifact cleanup, model-sync commits). Those are being dismissed separately with "won't fix" + per-alert justification — Scorecard penalizes any write, including the unavoidable kind.
Relates to #414
GMI Cloud is a HackwithBay 2.0 co-sponsor providing 40+ models (DeepSeek, Qwen, MiniMax, Gemini, Claude) via OpenAI-compatible API.
The existing
llm_openainode already works with GMI Cloud viabase_url_env— this PR documents that integration so hackathon participants can use it.Adds
docs/providers/gmi-cloud.mdwith:llm_openainodeSummary by CodeRabbit