Repository navigation
Docs audit: discovery, assistant guidance, and LFM3 recommendations - #172
Conversation
Model pages such as LFM3-1.7B are not in the docs.json navigation, so Mintlify left them out of site search and the docs assistant. In the last 90 days, no assistant answer cited an individual model page. Setting seo.indexing to "all" includes them. Pages with noindex: true stay excluded.
The box was hidden in #148, leaving Cmd/Ctrl+I as the only way to open the assistant on desktop. Users don't know the shortcut, so restore the visible entry point.
The context-length FAQ now names LFM3-1.7B and LFM3-VL-2B instead of the deprecated LFM2.5-2.6B. The two examples built on LFM2.5-Audio-1.5B get a note pointing to LFM3-Audio-2B. Absolute docs.liquid.ai links in examples are now root-relative.
Add 8 redirects for URLs that don't resolve but get real visitors, including links from Hugging Face and external articles to pages removed in earlier restructures.
Point speed, memory and device questions to Pipette and Hugging Face, benchmark questions to the model cards, and give a set answer to "which model powers you?" that links the Playground and API quickstart.
Describe LFM2.5-Embedding-350M as a text embedding model for retrieval in RAG pipelines, say that LFM2.5-ColBERT-350M is the reranking model, and name retrieval for RAG on the Nanos, Welcome and Complete Library pages so people searching for embeddings find these models.
|
Preview deployment for your docs. Learn more about Mintlify Previews.
💡 Tip: Enable Automations to automatically generate PRs for you. |
| - /lfm/models/lfm25-vl-3b | ||
| - /lfm/models/lfm25-vl-450m | ||
| - /lfm/models/lfm25-vl-450m-extract | ||
| - /lfm/models/lfm3-1-7b |
There was a problem hiding this comment.
Isn't this line duplicated?
There was a problem hiding this comment.
Not a duplicate: this entry is /lfm/models/lfm3-1-7b (hyphen), a mistyped URL that still gets visits; the next line is the real /lfm/models/lfm3-1.7b. The new redirect in docs.json sends the hyphen version to the real page, and the snapshot records it. No change.
🤖 Addressed by Claude Code
| - /lfm/fine-tuning/overview | ||
| - /lfm/fine-tuning/trl | ||
| - /lfm/fine-tuning/unsloth | ||
| - /lfm/frameworks/leap |
There was a problem hiding this comment.
Are we sure we want to reintroduce it?
There was a problem hiding this comment.
This does not bring the page back. /lfm/frameworks/leap was dropped in the #96 restructure with no redirect, but people still land on it from old links. This PR only adds a redirect from it to the "Migrating from LEAP SDK" page, which is the right place for anyone looking for LEAP now. The snapshot entry only records that redirect. Happy to drop it if you would rather that URL 404.
🤖 Addressed by Claude Code
alay2shah
changed the title
alay2shah
left a comment
alay2shah
left a comment
There was a problem hiding this comment.
Reviewed the final audit changes, including indexing exclusions, assistant source guidance, LFM3 recommendations, direct deprecation replacements, and verified ONNX links. Mintlify build validation and both GitHub checks passed.
Summary
Model pages outside the navigation were missing from search and assistant results, and general recommendations still led with older models. This audit makes model documentation discoverable and puts LFM3 first in general guidance while retaining useful older models.
.mintignoreand mark the contributing pagenoindex.qwen3_coderparsing from LFM2/LFM2.5 formats andlfm2parsing. Scope chat templates, LoRA module examples, generation settings, and context limits to the selected model.Validation
1b21fbe.Removing the assistant CSS restriction does not establish that Mintlify has enabled chat permissions; project configuration still controls assistant availability.
Preview review
Related work
A three-way check against #179 found two small conflicts in
guides/migration-guide.mdx: the shared tool-call snippet import and component invocation. When integrating #179, preserve this PR's LFM3 guidance, adopt itsLFM25ToolCallExamplename with paired JSX tags, and keep the unusedChatMLExampleimport removed.