Repository navigation
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
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[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 6095b18341
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".
|
|
||
| LibreChat has two separate kinds of search: | ||
|
|
||
| - **Search the web**: Turn on web search from the chat input, or give an agent the `web_search` capability, so the AI can look up current information. It needs a search provider, set up by your admin or with your own API keys when prompted. See [Web Search](/docs/features/web_search). |
There was a problem hiding this comment.
When an admin relies on this prerequisite summary and configures only a search provider—for example, Serper or SearXNG—web search still cannot run because a scraper is also mandatory. The linked guide explicitly requires both a search provider and scraper in content/docs/features/web_search.mdx lines 10–29 and 222, so this overview should mention both rather than implying that the provider alone is sufficient.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
|
|
||
| LibreChat has two separate kinds of search: | ||
|
|
||
| - **Search the web**: Turn on web search from the chat input, or give an agent the `web_search` capability, so the AI can look up current information. It needs a search provider, set up by your admin or with your own API keys when prompted. See [Web Search](/docs/features/web_search). |
There was a problem hiding this comment.
On deployments that customize endpoints.agents.capabilities and omit web_search, the web-search control is absent from regular chats because this capability gates all LLM interactions, not only individual Agents. Presenting the chat-input control and the capability as alternatives therefore misdirects users who cannot find the control; clarify that web_search must be allowed globally for the composer option, while an individual Agent enables Web Search through Agent Builder, as documented in the official Web Search availability conditions.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Fixed in 38991ba. The composer path (the + palette in a regular chat) and the Agent Builder path are now separate sentences, and a missing option links to features/web_search#web-search-option-is-not-visible, which covers the endpoints.agents.capabilities gate along with the role permission, endpoint and hideBadgeRow conditions. Covered by @Scenario:ai-overview-search-links-resolve, which clicks through to that anchor.
📦 Next.js Bundle Analysis for librechat.aiThis analysis was generated by the Next.js Bundle Analysis action. 🤖 This PR introduced no changes to the JavaScript bundle! 🙌 |
6095b18 to
3b2664c
Compare
📦 Next.js Bundle Analysis for librechat.aiThis analysis was generated by the Next.js Bundle Analysis action. 🤖 This PR introduced no changes to the JavaScript bundle! 🙌 |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 098b72b17d
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".
|
|
||
| ## Searching | ||
|
|
||
| LibreChat has two separate kinds of search: |
There was a problem hiding this comment.
For readers asking how to search uploaded documents, this categorical statement is inaccurate: LibreChat also exposes File Search, which content/docs/features/agents.mdx lines 125–132 describes as semantic search across uploaded documents, and the linked Message Search page likewise distinguishes that feature at lines 23–27. As written, the overview directs those users only to web and conversation-history search, so either add File Search or qualify these as the two kinds covered here.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Fixed in 29bd6be. The section now has a File Search bullet (the + palette in a regular chat, the File Search capability in the Agent Builder, the RAG API prerequisite, and a link to features/rag_api), and the lead-in no longer claims a fixed number of search kinds. Covered by @Scenario:ai-overview-file-search-explained in e2e/ai-overview-search-verify.spec.ts.
📦 Next.js Bundle Analysis for librechat.aiThis analysis was generated by the Next.js Bundle Analysis action. 🤖 This PR introduced no changes to the JavaScript bundle! 🙌 |
Discord feedback 1561862689301467167 asked how to search from the AI Overview page, which had no pointer to either search feature.
Fumadocs headings carry a Copy Anchor Link button, so the exact accessible name never equals the heading text.
098b72b to
29bd6be
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 29bd6be2e1
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".
|
|
||
| - **Search the web**: Let the AI look up current information. In a regular chat, open the **+** palette in the [message composer](/docs/features/composer) and select **Web Search**. For an agent, enable the **Web Search** capability in the Agent Builder instead. Web search needs both a search provider and a scraper, configured by your admin or with your own API keys when prompted. If the option is missing, see [Web Search option is not visible](/docs/features/web_search#web-search-option-is-not-visible). | ||
|
|
||
| - **Search your files**: Ask questions about documents you upload. In a regular chat, select **File Search** from the **+** palette. For an agent, enable the **File Search** capability in the Agent Builder. File Search finds passages by meaning and relies on the RAG API, which your admin sets up. See [RAG API (Chat with Files)](/docs/features/rag_api). |
There was a problem hiding this comment.
On deployments that omit file_search from a custom endpoints.agents.capabilities list or disable the FILE_SEARCH role permission, the palette and Agent Builder controls described here are unavailable even when the RAG API is healthy. The prerequisite sentence currently presents RAG setup as sufficient and, unlike the Web Search bullet, provides no missing-option guidance; content/docs/features/projects.mdx separately confirms that File Search must be enabled for the Agents endpoint, while the interface configuration documents the role gate. Add these availability conditions or link to guidance that explains them.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Fixed in d55df01, together with the shared cause behind the earlier web search findings: the overview was restating per-feature prerequisites, and every partial list left a gate out. The bullets now say only which backend each feature uses (provider and scraper, RAG API). One paragraph after the list says Web Search and File Search appear only when the deployment enables them, including the web_search/file_search agent capability and the matching role permission (checked against usePaletteEntries.tsx and useAgentCapabilities.ts on LibreChat dev), and links to the Web Search troubleshooting section and interface#filesearch for the full conditions. Covered by @Scenario:ai-overview-missing-search-option-explained.
📦 Next.js Bundle Analysis for librechat.aiThis analysis was generated by the Next.js Bundle Analysis action. 🤖 This PR introduced no changes to the JavaScript bundle! 🙌 |
List the capability and role-permission gates for Web Search and File Search in one non-exhaustive paragraph that links to the pages owning the full conditions.
📦 Next.js Bundle Analysis for librechat.aiThis analysis was generated by the Next.js Bundle Analysis action. 🤖 This PR introduced no changes to the JavaScript bundle! 🙌 |
This branch was successfully deployed
1 participant
Summary
The AI Overview page is linked as "Start Using LibreChat" from the user guides index, but it had no pointer to search. A reader on the Turkish version asked how to search. This adds a Searching section covering web search (the composer's + palette or an agent's Web Search capability, which needs a search provider and a scraper), File Search over uploaded documents (backed by the RAG API), and conversation search (the sidebar box, shown only when Meilisearch is enabled). One paragraph then explains that Web Search and File Search appear only when the deployment enables them, including the agent capability and the role permission, and links to the pages that hold the full conditions. I checked the gates against LibreChat
dev(ConversationsSection.tsx,usePaletteEntries.tsx,useAgentCapabilities.ts). Only the English source is edited.Change Type
Feedback addressed
1561862689301467167,/tr/docs/user_guides/ai_overviewTesting
e2e/ai-overview-search-verify.spec.ts(5 scenarios): passed in the CI End-to-end job (70 passed)npx prettier --checkandnpx eslinton the changed files: passedpnpm test: 38 files, 436 tests passednext devruns out of heap compiling the docs route on this machine, so the CI production-build job is the evidenceNot changed
1562012787734220814,/docs/features/ocr: a "good" rating with no message. Non-actionable.Checklist