update-models-ollama

Update src/modules/llms/server/ollama/ollama.models.ts with latest model definitions.

Reference src/modules/llms/server/llm.server.types.ts and src/modules/llms/server/models.mappings.ts for context only. Focus on the model file, do not descend into other code.

Automated Workflow:

# 1. Fetch the HTML to a cross-platform temp path (sorted by newest for stable ordering)
curl -s "https://ollama.com/library?sort=newest" -o "$(node -p "require('os').tmpdir()")/ollama-newest.html"

# 2. Parse it with the script (auto-finds the file in os.tmpdir())
node .claude/scripts/parse-ollama-models.js 2>&1

The parser outputs 5 pipe-delimited fields: modelName|pulls|capabilities|sizes|cloud

• deepseek-r1|91500000|tools,thinking|1.5b,7b,8b,14b,32b,70b,671b|
• kimi-k3|39000|vision,tools,thinking||cloud - cloud-only, so no sizes

Primary Sources:

• Model Library: https://ollama.com/library?sort=newest
• Parser script: .claude/scripts/parse-ollama-models.js

Fallbacks if blocked: Check https://github.com/ollama/ollama, search "ollama featured models", "ollama latest models", or search GitHub for latest model info

What the file is: a full mirror of the library index minus embedding models, in page order (newest first). ~222 entries as of 2026-08-17. The parser applies no pull threshold and no top-N cut: every line it prints belongs in the file, and every file entry should be a printed line.

Cloud models: cloud-only entries ARE carried (kimi-k3, glm-5.2, minimax-m3, ...). The 5th field is informational only - the file has no cloud marker. Empty sizes on a cloud line is expected, not a parse failure.

Removals: remove an entry only when its name is absent from the library index. Do NOT probe https://ollama.com/library/<id>: delisted models keep returning 200 there (verified: glm-5, glm-4.7 return 200 while absent from the index), so a 404 test never fires.

Sanity check the parse before editing: 200+ lines, non-zero pulls, capabilities present. All-zero pulls or empty capabilities means the page markup changed - fix the regexes in the parser (it warns on this), never write a stripped list into the file.

Field mapping:

• tools -> hasTools: true, vision -> hasVision: true, embedding -> isEmbeddings: true (note the 's'; the parser drops embedding models, so this should not come up)
• thinking and audio -> tags, followed by the size chips (1.5b, 7b, 32b, ...) in page order

Important:

• Sort them in the EXACT same order as the source (newest first, for stable ordering)
• Set today's date (YYYYMMDD format) for newly added models only
• Set OLLAMA_PREV_UPDATE to the previous OLLAMA_LAST_UPDATE, then OLLAMA_LAST_UPDATE to today (PREV drives the isNew badge)
• Do NOT change dates of existing models
• Review the full model list for additions, removals, and changes
• Minimize whitespace/comment changes, focus on content
• Preserve comments and newlines to make diffs easy to review