Skip to content

🤖 docs: Update the Endpoint and Agent librechat.yaml Reference - #863

Merged
berry-13 merged 3 commits into
mainfrom
berry-13/docs-yaml-endpoints-audit
Oct 23, 2026
Merged

berry-13 merged 3 commits into
mainfrom
berry-13/docs-yaml-endpoints-audit

Conversation

@berry-13

Copy link
Copy Markdown
Collaborator

Summary

Checks the endpoint and agent half of the librechat.yaml reference (agents, summarization, Azure OpenAI, Anthropic on Vertex, Bedrock, assistants, custom endpoints and parameters, model config and specs, shared endpoint settings) and the AI endpoint guides against the LibreChat config schemas. It fixes wrong keys and defaults, removes keys the server no longer reads, and documents the missing ones.

Stacked on #862; review only this PR's diff.

Change Type

  • Documentation update

Testing

  • Each changed key, type and default traced to packages/data-provider/src and its consumers on LibreChat dev, then rechecked by a second independent pass
  • Internal links and anchors: no new broken links
  • Frontmatter parses

Checklist

  • No dash punctuation added
  • OptionTable cells are plain text

@vercel

vercel Bot commented Oct 23, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
librechat-ai Ready Ready Preview Oct 23, 2026 8:42pm UTC

Request Review

@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Oct 23, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-10-23T20:36:19.241719Z c491b49 New commits
🔒 Security Review ✅ Completed 2026-10-23T15:20:10.728088Z 2b11f3c PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2b11f3c8bb

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

]}
/>

**Default:** `2` for `assistants`, `1` for `azureAssistants`

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

P2 Badge Correct the Azure Assistants parsed default

Clarify that an existing azureAssistants block does not currently default to version 1. Both assistants and azureAssistants use assistantEndpointSchema, whose version field has a schema default of 2; therefore adding an Azure block for another setting while omitting version produces version 2, and the endpoint-specific fallback of 1 is only reachable when no parsed version exists. Users relying on this statement must explicitly set version: 1.

Useful? React with 👍 / 👎.

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

Partly right, fixed in 9ba7376: assistantsConfigSetup keeps the default version 1 only when endpoints.azureOpenAI.assistants loaded the Azure defaults (packages/data-schemas/src/app/endpoints.ts:44-53); otherwise it is 2. Both the table and the Default line now say so.

'maxEnvironmentChoices',
'Integer',
'Maximum additional machine choices saved on one agent (1-128).',
'Unset',

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

P2 Badge State the effective environment-choice default

Replace Unset with the effective default of 32. The same release's configuration changelog explicitly says maxEnvironmentChoices defaults to 32 with a ceiling of 128, so this reference currently implies that omitting the field leaves no configured limit even though users will encounter the 32-choice cap.

Useful? React with 👍 / 👎.

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

Fixed in 9ba7376: the default is 32 (DEFAULT_AGENT_CODE_ENVIRONMENT_CHOICES).


**Example:**
```yaml filename="endpoints / azureOpenAI / groups / {group_item} / models / {model_item=Object} / assistants"
assistants: true

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

P2 Badge Remove the ignored model-level assistants example

Do not demonstrate assistants: true on a model object. Although the schema accepts this property, Azure Assistant models are selected only from groups whose group-level assistants flag is enabled, as the description immediately above already notes. A user copying this example while leaving the group flag unset gets no Azure Assistant models, so move the example to the group-level section or remove it.

Useful? React with 👍 / 👎.

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

Fixed in 9ba7376: the example now shows the flag on the group, where it takes effect.

@berry-13
berry-13 added this pull request to stack #869 October 23, 2026 16:25
@berry-13
berry-13 force-pushed the berry-13/docs-yaml-endpoints-audit branch 3 times, most recently from 35af0fd to c472c91 Compare October 23, 2026 19:39
Base automatically changed from berry-13/docs-yaml-core-audit to main October 23, 2026 20:34
Check the agents, summarization, Azure, Vertex, Bedrock, assistants,
custom endpoint, model spec and shared endpoint pages, plus the AI
endpoint guides, against the LibreChat config schemas: fix wrong keys
and defaults, remove keys the server no longer reads and document the
missing ones.
Protocol version 1 is advertised when CODE_ENVIRONMENT_DECISION_VERSION
is unset, so a rolling upgrade must set it to 0 before the first new
replica starts and remove it once every replica is upgraded.
…ants flag

azureAssistants falls back to version 1 only when
endpoints.azureOpenAI.assistants is enabled. maxEnvironmentChoices
defaults to 32. Show the assistants flag on the group, where it takes
effect, rather than on a model.
@berry-13
berry-13 force-pushed the berry-13/docs-yaml-endpoints-audit branch from c472c91 to c491b49 Compare October 23, 2026 20:34
@github-actions

Copy link
Copy Markdown
Contributor

📦 Next.js Bundle Analysis for librechat.ai

This analysis was generated by the Next.js Bundle Analysis action. 🤖

This PR introduced no changes to the JavaScript bundle! 🙌

@berry-13
berry-13 merged commit f7042a4 into main Oct 23, 2026
5 checks passed
@berry-13
berry-13 deleted the berry-13/docs-yaml-endpoints-audit branch October 23, 2026 20:56

This branch was successfully deployed

1 active deployment
Preview — c491b49f Deployed Oct 23, 2026 by vercel[bot]
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.

1 participant