clawdbot/docs/models.md

summary, read_when

summary	read_when
Plan for models CLI: scan, list, aliases, fallbacks, status	Adding or modifying models CLI (models list/set/scan/aliases/fallbacks) Changing model fallback behavior or selection UX Updating model scan probes (tools/images)

Models CLI plan

Goal: give clear model visibility + control (configured vs available), plus scan tooling that prefers tool-call + image-capable models and maintains ordered fallbacks.

Command tree (draft)

• clawdbot models list
  • default: configured models only
  • flags: --all (full catalog), --local, --provider <name>, --json, --plain
• clawdbot models status
  • show default model + aliases + fallbacks + allowlist
• clawdbot models set <modelOrAlias>
  • writes agent.model in config
• clawdbot models set-image <modelOrAlias>
  • writes agent.imageModel in config
• clawdbot models aliases list|add|remove
  • writes agent.modelAliases
• clawdbot models fallbacks list|add|remove|clear
  • writes agent.modelFallbacks
• clawdbot models image-fallbacks list|add|remove|clear
  • writes agent.imageModelFallbacks
• clawdbot models scan
  • OpenRouter :free scan; probe tool-call + image; interactive selection

Config changes

• Add agent.modelFallbacks: string[] (ordered list of provider/model IDs).
• Add agent.imageModel?: string (optional image-capable model for image tool).
• Add agent.imageModelFallbacks?: string[] (ordered list for image tool).
• Keep existing:
  • agent.model (default)
  • agent.allowedModels (list filter)
  • agent.modelAliases (shortcut names)

Scan behavior (models scan)

Input

• OpenRouter /models list (filter :free)
• Requires OPENROUTER_API_KEY (or stored OpenRouter key in auth storage)
• Optional filters: --max-age-days, --min-params, --provider, --max-candidates
• Probe controls: --timeout, --concurrency

Probes (direct pi-ai complete)

• Tool-call probe (required):
  • Provide a dummy tool, verify tool call emitted.
• Image probe (preferred):
  • Prompt includes 1x1 PNG; success if no "unsupported image" error.

Scoring/selection

• Prefer models passing tool + image for text/tool fallbacks.
• Prefer image-only models for image tool fallback (even if tool probe fails).
• Rank by: image ok, then lower tool latency, then larger context, then params.

Interactive selection (TTY)

• Multiselect list with per-model stats:
  • model id, tool ok, image ok, median latency, context, inferred params.
• Pre-select top N (default 6).
• Non-TTY: auto-select; require --yes/--no-input to apply.

Output

• Writes agent.modelFallbacks ordered.
• Writes agent.imageModelFallbacks ordered (image-capable models).
• Optional --set-default to set agent.model.
• Optional --set-image to set agent.imageModel.

Runtime fallback

• On model failure: try agent.modelFallbacks in order.
• Ignore fallback entries not in agent.allowedModels (if allowlist set).
• Persist last successful provider/model to session entry.
• /status shows last used model (not just default).

Tests

• Unit: scan selection ordering + probe classification.
• CLI: list/aliases/fallbacks add/remove + scan writes config.
• Status: shows last used model + fallbacks.

Docs

• Update docs/configuration.md with agent.modelFallbacks.
• Keep this doc current when CLI surface or scan logic changes.