Merge pull request #4 from DrJLabs/feat/profile-instance-management

feat: add registry-backed instance management previews

Author: DrJsPBs

DOCKER.md

@@ -20,6 +20,14 @@
 
 The compose stack now mounts the whole `./prompts` tree into `/app/prompts` and each service keeps its own runtime prompt config file under `/data`.
 
+Registry-backed management metadata now also lives in:
+
+- `config/profiles/*.yaml`
+- `config/instances/*.yaml`
+
+Those registry files are used for validation, instance listing, and runtime previews.
+They do not replace the live prompt manager or the live `docker-compose.yml` entrypoint in this phase.
+
 Default runtime prompt selection:
 
 - `chatmock` uses `${CHATMOCK_PROMPT_DIR:-/app/prompts/bare}`
@@ -72,6 +80,24 @@ curl -sS -X POST http://127.0.0.1:8000/admin/prompts/reload \
   -H "X-ChatMock-Admin-Token: $CHATMOCK_ADMIN_TOKEN" | jq .
 ```
 
+## Instance registry inspection
+
+Registry-backed read/preview controls are also exposed locally:
+
+- `GET /admin/profiles`
+- `GET /admin/instances`
+- `GET /admin/instances/<instance_id>/preview`
+
+CLI equivalents:
+
+```bash
+chatmock instances list
+chatmock instances validate
+chatmock instances preview chatmock
+```
+
+These controls are intended to give the current GUI or a future replacement UI a stable backend contract before any lifecycle/provisioning work is added.
+
 ## Isolated test stack
 
 Use the override file to bring up a disposable stack on separate ports without touching the live services or their `/data` volume.

chatmock.py

@@ -1,7 +1,8 @@
 from __future__ import annotations
 
+import sys
+
 from chatmock.cli import main
 
 if __name__ == "__main__":
-    main()
-
+    sys.exit(main())

chatmock/app.py

@@ -10,6 +10,7 @@
 
 from .config import get_prompt_manager
 from .http import build_cors_headers
+from .instance_service import build_instance_service
 from .routes_openai import openai_bp
 from .routes_ollama import ollama_bp
 from .websocket_routes import register_websocket_routes
@@ -49,6 +50,7 @@ def create_app(
         DEFAULT_WEB_SEARCH=bool(default_web_search),
         INJECT_DEFAULT_INSTRUCTIONS=bool(inject_default_instructions),
         PROMPT_MANAGER=prompt_manager,
+        INSTANCE_SERVICE=None,
         ADMIN_TOKEN=(
             admin_token
             if isinstance(admin_token, str) and admin_token
@@ -107,6 +109,16 @@ def _is_in_trusted_ranges(value: str | None) -> bool:
                 return jsonify({"error": {"message": "Invalid admin token"}}), 403
         return None
 
+    def _get_instance_service():
+        service = app.config.get("INSTANCE_SERVICE")
+        if service is None:
+            service = build_instance_service()
+            app.config["INSTANCE_SERVICE"] = service
+        return service
+
+    def _registry_error(exc: Exception):
+        return jsonify({"error": {"message": str(exc)}}), 400
+
     @app.get("/admin/prompts")
     def admin_prompts_state():
         denied = _require_local_admin()
@@ -142,6 +154,43 @@ def admin_prompts_config():
             return jsonify({"error": {"message": str(exc)}}), 400
         return jsonify(prompt_manager.as_dict())
 
+    @app.get("/admin/profiles")
+    def admin_profiles():
+        denied = _require_local_admin()
+        if denied is not None:
+            return denied
+        try:
+            service = _get_instance_service()
+        except (FileNotFoundError, OSError, ValueError) as exc:
+            return _registry_error(exc)
+        return jsonify({"profiles": service.list_profiles()})
+
+    @app.get("/admin/instances")
+    def admin_instances():
+        denied = _require_local_admin()
+        if denied is not None:
+            return denied
+        try:
+            service = _get_instance_service()
+        except (FileNotFoundError, OSError, ValueError) as exc:
+            return _registry_error(exc)
+        return jsonify({"instances": service.list_instances()})
+
+    @app.get("/admin/instances/<instance_id>/preview")
+    def admin_instance_preview(instance_id: str):
+        denied = _require_local_admin()
+        if denied is not None:
+            return denied
+        try:
+            service = _get_instance_service()
+        except (FileNotFoundError, OSError, ValueError) as exc:
+            return _registry_error(exc)
+        try:
+            preview = service.render_instance_preview(instance_id)
+        except ValueError as exc:
+            return jsonify({"error": {"message": str(exc)}}), 404
+        return jsonify(preview)
+
     @app.after_request
     def _cors(resp):
         for k, v in build_cors_headers().items():

chatmock/bundled_config/instances/chatmock-clawmem.yaml

@@ -0,0 +1,20 @@
+id: chatmock-clawmem
+label: ChatMock ClawMem
+profile_id: clawmem
+bind_host: 127.0.0.1
+port: 8001
+runtime: docker_compose
+prompt_config_path: /data/prompt-config-chatmock-clawmem.json
+state_group: shared-auth-default
+compose_service_name: chatmock-clawmem
+container_name: chatmock-clawmem
+env_prefix: CHATMOCK_CLAWMEM
+env_overrides: {}
+healthcheck:
+  path: /health
+ui:
+  order: 20
+  mutable_fields:
+    - profile_id
+    - port
+enabled: true

chatmock/bundled_config/instances/chatmock.yaml

@@ -0,0 +1,20 @@
+id: chatmock
+label: ChatMock
+profile_id: bare
+bind_host: 127.0.0.1
+port: 8000
+runtime: docker_compose
+prompt_config_path: /data/prompt-config-chatmock.json
+state_group: shared-auth-default
+compose_service_name: chatmock
+container_name: chatmock
+env_prefix: CHATMOCK
+env_overrides: {}
+healthcheck:
+  path: /health
+ui:
+  order: 10
+  mutable_fields:
+    - profile_id
+    - port
+enabled: true

chatmock/bundled_config/profiles/bare.yaml

@@ -0,0 +1,11 @@
+id: bare
+label: Bare
+description: Default low-opinion prompt set for the main ChatMock service
+prompt_dir: chatmock/bundled_prompts/bare
+base_prompt_file: prompt.md
+codex_prompt_file: prompt_gpt5_codex.md
+runtime_defaults:
+  inject_default_instructions: true
+ui:
+  order: 10
+  editable: true

chatmock/bundled_config/profiles/clawmem.yaml

@@ -0,0 +1,11 @@
+id: clawmem
+label: ClawMem
+description: ClawMem-specific prompt set for the secondary managed service
+prompt_dir: chatmock/bundled_prompts/clawmem
+base_prompt_file: prompt.md
+codex_prompt_file: prompt_gpt5_codex.md
+runtime_defaults:
+  inject_default_instructions: true
+ui:
+  order: 20
+  editable: true

chatmock/bundled_prompts/bare/prompt.md

@@ -0,0 +1,5 @@
+Follow the developer and user instructions contained in the input.
+When calling tools, you must strictly match the provided tool schema.
+Do not call a tool with empty arguments unless the schema explicitly allows no arguments.
+Always include every required argument before making the tool call.
+If a tool call returns a schema error, inspect the schema and try again with corrected arguments.

chatmock/bundled_prompts/bare/prompt_gpt5_codex.md

@@ -0,0 +1,5 @@
+Follow the developer and user instructions contained in the input.
+When calling tools, you must strictly match the provided tool schema.
+Do not call a tool with empty arguments unless the schema explicitly allows no arguments.
+Always include every required argument before making the tool call.
+If a tool call returns a schema error, inspect the schema and try again with corrected arguments.

chatmock/bundled_prompts/clawmem/prompt.md

@@ -0,0 +1,25 @@
+You are a structured memory-enrichment model for ClawMem.
+
+Your outputs feed a long-lived memory system. Precision, consistency, and schema discipline matter more than creativity or recall.
+
+You may be asked to do one of four tasks:
+1. construct a memory note
+2. extract named entities
+3. classify document-to-document relations
+4. decide whether a memory should evolve
+
+Global rules:
+- Follow the task prompt exactly.
+- If the task asks for JSON, return valid JSON only.
+- Do not add prose, markdown fences, headings, or explanations outside the requested JSON.
+- Never copy placeholder examples from the prompt.
+- Prefer omission over invention.
+- If uncertain, return fewer items.
+- Be conservative, stable, and repeatable.
+
+Schema discipline:
+- Use exactly the requested keys and value types.
+- Do not add extra keys.
+- Do not rename keys.
+- Do not emit partial JSON.
+- If the best answer is "nothing," still return valid empty JSON in the requested schema.