feat: add unload on_agent_switch builtin hook

agent-schema.json

@@ -180,6 +180,15 @@
             "ANTHROPIC_API_KEY"
           ]
         },
+        "unload_api": {
+          "type": "string",
+          "description": "Optional path (or absolute URL) to the provider's model-unload endpoint. POSTed with `{\"model\": \"<id>\"}` when the agent wires the `unload` builtin into its `on_agent_switch` hook chain, to free GPU/RAM held by the previous model. Today only Docker Model Runner ships a provider that calls this endpoint; cloud providers don't implement [provider.Unloader] and the hook silently skips them. A relative path is resolved against the scheme+host of base_url; an absolute URL is used verbatim.",
+          "examples": [
+            "/engines/_unload",
+            "/api/unload",
+            "http://localhost:11434/api/unload"
+          ]
+        },
         "temperature": {
           "type": "number",
           "description": "Default sampling temperature for models using this provider.",
@@ -872,7 +881,7 @@
         },
         "type": {
           "type": "string",
-          "description": "Type of hook. 'command' executes a shell command; 'builtin' invokes a named in-process Go function registered by the runtime; 'model' asks an LLM and translates its reply into the hook's native output (used for LLM-as-a-judge pre_tool_use, summarizers, etc., with no Go code). The docker-agent runtime ships these builtins: 'add_date' (turn_start: today's date), 'add_environment_info' (session_start: cwd, git, OS, arch), 'add_prompt_files' (turn_start: contents of named files looked up in the workdir hierarchy and the home directory), 'add_git_status' (turn_start: `git status --short --branch`), 'add_git_diff' (turn_start: `git diff --stat`, or full diff with args=['full']), 'add_directory_listing' (session_start: top-level entries of cwd), 'add_user_info' (session_start: current OS user and hostname), 'add_recent_commits' (session_start: `git log --oneline -n N`, default N=10, override via args=['<N>']), 'max_iterations' (before_llm_call: hard stop after N model calls; args=['<N>'] required), 'redact_secrets' (pre_tool_use / before_llm_call / tool_response_transform: scrubs detected secrets from tool arguments, outgoing chat content, and tool output — the same builtin handles all three legs and dispatches on the event; the matching agent-level 'redact_secrets: true' flag auto-injects the entries for all three).",
+          "description": "Type of hook. 'command' executes a shell command; 'builtin' invokes a named in-process Go function registered by the runtime; 'model' asks an LLM and translates its reply into the hook's native output (used for LLM-as-a-judge pre_tool_use, summarizers, etc., with no Go code). The docker-agent runtime ships these builtins: 'add_date' (turn_start: today's date), 'add_environment_info' (session_start: cwd, git, OS, arch), 'add_prompt_files' (turn_start: contents of named files looked up in the workdir hierarchy and the home directory), 'add_git_status' (turn_start: `git status --short --branch`), 'add_git_diff' (turn_start: `git diff --stat`, or full diff with args=['full']), 'add_directory_listing' (session_start: top-level entries of cwd), 'add_user_info' (session_start: current OS user and hostname), 'add_recent_commits' (session_start: `git log --oneline -n N`, default N=10, override via args=['<N>']), 'max_iterations' (before_llm_call: hard stop after N model calls; args=['<N>'] required), 'redact_secrets' (pre_tool_use / before_llm_call / tool_response_transform: scrubs detected secrets from tool arguments, outgoing chat content, and tool output — the same builtin handles all three legs and dispatches on the event; the matching agent-level 'redact_secrets: true' flag auto-injects the entries for all three), 'unload' (on_agent_switch: walks the previous agent's models and calls Unload() on every provider that implements provider.Unloader — e.g. asks Docker Model Runner to release the GPU/RAM held by the just-departing model so the next agent's model can claim it; opt in by adding the entry to the agent's hooks.on_agent_switch list).",
           "enum": [
             "command",
             "builtin",
@@ -1012,7 +1021,7 @@
         },
         "provider_opts": {
           "type": "object",
-          "description": "Provider-specific options. Sampling parameters: top_k (integer, supported by anthropic, google, amazon-bedrock, and custom OpenAI-compatible providers like vLLM/Ollama), repetition_penalty (float, forwarded to custom OpenAI-compatible providers), min_p (float, forwarded to custom providers), seed (integer, forwarded to OpenAI). Infrastructure options: http_headers (map of string to string, adds custom HTTP headers to every request; used for OpenAI-compatible providers like github-copilot which requires Copilot-Integration-Id). dmr: runtime_flags. anthropic/amazon-bedrock (Claude): interleaved_thinking (boolean, default true), thinking_display ('summarized', 'omitted', or 'display') controls whether thinking blocks are returned in responses when thinking is enabled. Claude Opus 4.7 hides thinking by default ('omitted'); set thinking_display: summarized (or thinking_display: display) to receive thinking blocks. openai: transport ('sse' or 'websocket') to choose between SSE and WebSocket streaming for the Responses API. openai/anthropic/google: rerank_prompt (string) to fully override the system prompt used for RAG reranking (advanced - prefer using results.reranking.criteria for domain-specific guidance). Google: google_search (boolean) enables Google Search grounding, google_maps (boolean) enables Google Maps grounding, code_execution (boolean) enables server-side code execution.",
+          "description": "Provider-specific options. Sampling parameters: top_k (integer, supported by anthropic, google, amazon-bedrock, and custom OpenAI-compatible providers like vLLM/Ollama), repetition_penalty (float, forwarded to custom OpenAI-compatible providers), min_p (float, forwarded to custom providers), seed (integer, forwarded to OpenAI). Lifecycle: unload_api (string) overrides the unload endpoint inherited from the provider config (relative path resolved against base_url's scheme+host, or an absolute URL); used by the runtime's `unload` on_agent_switch builtin hook to release model resources between agent switches. Infrastructure options: http_headers (map of string to string, adds custom HTTP headers to every request; used for OpenAI-compatible providers like github-copilot which requires Copilot-Integration-Id). dmr: runtime_flags. anthropic/amazon-bedrock (Claude): interleaved_thinking (boolean, default true), thinking_display ('summarized', 'omitted', or 'display') controls whether thinking blocks are returned in responses when thinking is enabled. Claude Opus 4.7 hides thinking by default ('omitted'); set thinking_display: summarized (or thinking_display: display) to receive thinking blocks. openai: transport ('sse' or 'websocket') to choose between SSE and WebSocket streaming for the Responses API. openai/anthropic/google: rerank_prompt (string) to fully override the system prompt used for RAG reranking (advanced - prefer using results.reranking.criteria for domain-specific guidance). Google: google_search (boolean) enables Google Search grounding, google_maps (boolean) enables Google Maps grounding, code_execution (boolean) enables server-side code execution.",
           "additionalProperties": true
         },
         "track_usage": {