feat: add GLM (Z.ai) LLM provider - #10216
Open
shawonis08 wants to merge 3 commits into
Open
Conversation
Add first-class GLM (Z.ai) provider support to pgAdmin AI configuration and LLM runtime
The GLM provider defaulted to https://open.bigmodel.cn/api/paas/v4, which is Zhipu's China platform, not Z.ai. A Z.ai API key sent there is rejected, so every AI assistant request failed with a 401. ALLOWED_LLM_API_URLS compounded this by allowlisting only open.bigmodel.cn, so overriding the URL in preferences was rejected as well. Default to https://api.z.ai/api/paas/v4 as documented at docs.z.ai, and allowlist api.z.ai alongside open.bigmodel.cn so the Z.ai open platform, Coding Plan subscriptions (https://api.z.ai/api/coding/paas/v4) and the China platform all work. Allowlist matching is scheme+host+port, so one api.z.ai entry covers both Z.ai paths. Also fixed while verifying against the live API: - Send max_tokens rather than the inherited max_completion_tokens. Z.ai ignores the latter outright, silently removing the output limit instead of failing. Overridden in GLMClient only; it is the sole subclass of OpenAIClient, so OpenAI and Docker keep the parameter they want. - Never fall back to the Responses API. Z.ai exposes no /responses endpoint, so an unrelated error matching the inherited heuristic would switch the client onto a path that cannot work and mask the real error. - Track reasoning_content when reading a Chat Completions stream. GLM-4.6+ returns chain-of-thought there, and a response that spends its whole budget reasoning has no content and no tool calls, which surfaced as a misleading "No response content returned from API". Report it as the token-limit error it is. OpenAI never sends this field, so its behaviour is unchanged. - Drop the hardcoded glm-4.5 default. The available models differ across the three platforms, and OpenAIClient's fallback meant an unconfigured GLM silently POSTed "model": "gpt-4o" to Z.ai and 404'd. get_llm_client() now raises a message naming the preference to set, and is_available() requires both a key and a model. Verified against the live Z.ai API: streaming, tool calling, usage reporting, the no-model guard and the reasoning-only path. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
WalkthroughGLM (Z.ai) is added as an OpenAI-compatible LLM provider with configuration, client creation, model refresh endpoints, preference integration, validation tests, and updated documentation. ChangesGLM Provider Support
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant PreferencesUI
participant refresh_glm_models
participant GLMModelFetcher
PreferencesUI->>refresh_glm_models: submit API URL and key file
refresh_glm_models->>refresh_glm_models: validate URL and key file
refresh_glm_models->>GLMModelFetcher: fetch GLM models
GLMModelFetcher-->>PreferencesUI: return models or error
Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ 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 |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 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 `@web/pgadmin/llm/__init__.py`:
- Around line 253-258: The GLM API URL guidance must explain that remote custom
endpoints require administrator allowlisting. Update the preference help text in
web/pgadmin/llm/__init__.py around the GLM URL setting to mention
ALLOWED_LLM_API_URLS and the required scheme, host, and port; also document this
configuration requirement for non-Z.ai custom URLs in docs/en_US/preferences.rst
around lines 86-89.
In `@web/pgadmin/llm/README.md`:
- Around line 23-27: Update the configuration list in the LLM README to include
docker among the supported DEFAULT_LLM_PROVIDER values and document
GLM_API_MODEL as required for GLM initialization, alongside GLM_API_KEY_FILE.
🪄 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: CHILL
Plan: Pro Plus
Run ID: 66dc7b2a-3621-4e6e-b03a-ca8b6769b4fd
📒 Files selected for processing (13)
docs/en_US/ai_tools.rstdocs/en_US/preferences.rstweb/config.pyweb/pgadmin/llm/README.mdweb/pgadmin/llm/__init__.pyweb/pgadmin/llm/client.pyweb/pgadmin/llm/providers/__init__.pyweb/pgadmin/llm/providers/glm.pyweb/pgadmin/llm/tests/test_api_url_validation.pyweb/pgadmin/llm/tests/test_llm_status.pyweb/pgadmin/llm/tests/test_model_refresh_endpoints.pyweb/pgadmin/llm/utils.pyweb/pgadmin/preferences/__init__.py
Comment on lines
+253
to
+258
| help_str=gettext( | ||
| 'URL for the GLM API endpoint. Leave empty to use ' | ||
| 'the default (https://api.z.ai/api/paas/v4). Use ' | ||
| 'https://api.z.ai/api/coding/paas/v4 for a Coding Plan ' | ||
| 'subscription, or set a custom URL to use another ' | ||
| 'GLM-compatible API provider.' |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
Explain the allowlist requirement for remote custom endpoints.
The default allowlist accepts api.z.ai and loopback hosts only, so a remote custom GLM-compatible URL will be rejected unless an administrator adds its scheme, host, and port to ALLOWED_LLM_API_URLS.
web/pgadmin/llm/__init__.py#L253-L258: State this constraint in the preference help text.docs/en_US/preferences.rst#L86-L89: Document the required administrator allowlist configuration for non-Z.ai custom URLs.
📍 Affects 2 files
web/pgadmin/llm/__init__.py#L253-L258(this comment)docs/en_US/preferences.rst#L86-L89
🤖 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 `@web/pgadmin/llm/__init__.py` around lines 253 - 258, The GLM API URL guidance
must explain that remote custom endpoints require administrator allowlisting.
Update the preference help text in web/pgadmin/llm/__init__.py around the GLM
URL setting to mention ALLOWED_LLM_API_URLS and the required scheme, host, and
port; also document this configuration requirement for non-Z.ai custom URLs in
docs/en_US/preferences.rst around lines 86-89.
Comment on lines
+23
to
27
| - `DEFAULT_LLM_PROVIDER`: Set to 'anthropic', 'openai', 'glm', or 'ollama' | ||
| - `ANTHROPIC_API_KEY_FILE`: Path to file containing Anthropic API key | ||
| - `OPENAI_API_KEY_FILE`: Path to file containing OpenAI API key | ||
| - `GLM_API_KEY_FILE`: Path to file containing GLM API key | ||
| - `OLLAMA_API_URL`: URL for Ollama server (e.g., 'http://localhost:11434') |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
Document the required GLM model and Docker provider.
Line 23 omits supported docker; Lines 24-27 omit GLM_API_MODEL, although GLM initialization fails until a model is selected. Add both to keep config-file setup usable.
🤖 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 `@web/pgadmin/llm/README.md` around lines 23 - 27, Update the configuration
list in the LLM README to include docker among the supported
DEFAULT_LLM_PROVIDER values and document GLM_API_MODEL as required for GLM
initialization, alongside GLM_API_KEY_FILE.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds GLM (Z.ai) as an LLM provider for pgAdmin's AI features, alongside
Anthropic, OpenAI, Ollama and Docker Model Runner. GLM exposes an
OpenAI-compatible API, so
GLMClientsubclassesOpenAIClientandoverrides only where Z.ai diverges.
Configuration
GLM_API_URL— defaults tohttps://api.z.ai/api/paas/v4GLM_API_KEY_FILE— defaults to~/.zai-api-keyGLM_API_MODEL— no default (see below)ALLOWED_LLM_API_URLSgainshttps://api.z.ai:443.Allowlist matching is scheme+host+port, so the single
api.z.aientry also covers Coding Plan subscriptions athttps://api.z.ai/api/coding/paas/v4.Testing
Verified against the live Z.ai API: streaming text, tool calling, usage
reporting, the no-model guard, and the reasoning-only path.
Unit tests cover URL allowlisting for all three platforms (including the
Coding Plan path), preference resolution and rejection, the model-refresh
endpoints, and LLM status reporting.
Docs updated in
ai_tools.rstandpreferences.rst.Summary by CodeRabbit
New Features
Documentation