Summary

Adds a local Docker dev environment with optional Ollama support while keeping external LLM providers easy to use for stronger model testing.

What changed

• Adds Ollama to the dev Compose stack with persistent model storage and a make dev-ollama-pull helper.
• Updates the dev hub config example and Docker docs for local-only Ollama testing and external-provider testing.
• Supports ollama across hub-side AI config, streaming, failure summaries, settings, and doctor checks.
• Uses OpenClaw native Ollama onboarding for Docker agents with http://ollama:11434 and strips the ollama/ model prefix.
• Stabilizes Docker agent bootstrap by waiting for workspace files before starting OpenClaw bootstrap work.
• Pins OpenClaw consistently and fails the agent image build if the pinned install fails.
• Adds focused tests for Ollama provider selection, doctor checks, bootstrap flags/config, Docker copy behavior, and settings display.

Validation

• docker run --rm -e GOFLAGS=-buildvcs=false -v "$PWD":/app -w /app golang:1.25 go test ./cmd/claw-bridge ./pkg/provider/docker ./pkg/hub ./pkg/types
• docker compose -f docker/compose.dev.yml config
• make dev-agent-build

npm run lint was also attempted, but it currently fails on existing repo-wide React hook lint errors unrelated to this change.

Linked issue

Closes #328