Add per-model council settings, Quick Chat council mode, and per-member trace selection

Council members now support individual temperature, reasoning effort, web search, and
context trace overrides. Quick Chat inherits council config from the source node and
streams through the 3-stage council pipeline. Blueprint loading migrates old string[]
council formats to CouncilMemberConfig[].

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

3 files changed, 40 insertions, 12 deletions

diff --git a/backend/app/main.py b/backend/app/main.py
index 9370a32..304c74f 100644
--- a/backend/app/main.py
+++ b/backend/app/main.py
@@ -472,12 +472,13 @@ async def run_council_stream(
 
     openrouter_key = get_user_api_key(resolved, "openrouter")
 
-    # Build LLMConfig + attachments for each council member
+    # Build LLMConfig + attachments + per-member contexts for each council member
     member_configs: list[LLMConfig] = []
     attachments_per_model: list[list[dict] | None] = []
     tools_per_model: list[list[dict] | None] = []
+    contexts_per_model: list[Context | None] = []
 
-    all_model_names = [m.model_name for m in request.council_models] + [request.chairman_model]
+    all_model_names = [m.model_name for m in request.council_models] + [request.chairman_model.model_name]
 
     for member in request.council_models:
         provider = resolve_provider(member.model_name)
@@ -487,10 +488,11 @@ async def run_council_stream(
         config = LLMConfig(
             provider=provider,
             model_name=member.model_name,
-            temperature=request.temperature,
+            temperature=member.temperature if member.temperature is not None else request.temperature,
             system_prompt=request.system_prompt,
             api_key=api_key,
-            reasoning_effort=request.reasoning_effort,
+            reasoning_effort=member.reasoning_effort if member.reasoning_effort is not None else request.reasoning_effort,
+            enable_google_search=member.enable_google_search if member.enable_google_search is not None else request.enable_google_search,
         )
         member_configs.append(config)
 
@@ -539,16 +541,31 @@ async def run_council_stream(
         attachments_per_model.append(attachments or None)
         tools_per_model.append(tools or None)
 
+        # Per-member context override
+        if member.incoming_contexts:
+            raw = []
+            for ctx in member.incoming_contexts:
+                raw.extend(ctx.messages)
+            if request.merge_strategy == MergeStrategy.SMART:
+                merged = smart_merge_messages(raw)
+            else:
+                merged = raw
+            contexts_per_model.append(Context(messages=merged))
+        else:
+            contexts_per_model.append(None)  # Use shared context
+
     # Build chairman config
-    chairman_provider = resolve_provider(request.chairman_model)
+    chairman = request.chairman_model
+    chairman_provider = resolve_provider(chairman.model_name)
     chairman_api_key = get_user_api_key(resolved, chairman_provider.value)
     chairman_config = LLMConfig(
         provider=chairman_provider,
-        model_name=request.chairman_model,
-        temperature=request.temperature,
+        model_name=chairman.model_name,
+        temperature=chairman.temperature if chairman.temperature is not None else request.temperature,
         system_prompt=request.system_prompt,
         api_key=chairman_api_key,
-        reasoning_effort=request.reasoning_effort,
+        reasoning_effort=chairman.reasoning_effort if chairman.reasoning_effort is not None else request.reasoning_effort,
+        enable_google_search=chairman.enable_google_search if chairman.enable_google_search is not None else request.enable_google_search,
     )
 
     return StreamingResponse(
@@ -561,6 +578,7 @@ async def run_council_stream(
             tools_per_model=tools_per_model,
             openrouter_api_key=openrouter_key,
             images=images,
+            contexts_per_model=contexts_per_model,
         ),
         media_type="text/event-stream",
     )
diff --git a/backend/app/schemas.py b/backend/app/schemas.py
index a527004..4213f15 100644
--- a/backend/app/schemas.py
+++ b/backend/app/schemas.py
@@ -57,16 +57,21 @@ class NodeRunRequest(BaseModel):
     
 class CouncilMemberConfig(BaseModel):
     model_name: str  # e.g. "gpt-5", "claude-opus-4-6", "gemini-3-pro-preview"
+    temperature: Optional[float] = None           # None = use request default
+    reasoning_effort: Optional[ReasoningEffort] = None
+    enable_google_search: Optional[bool] = None
+    incoming_contexts: Optional[List[Context]] = None  # Per-member context override
 
 class CouncilRunRequest(BaseModel):
     node_id: str
-    incoming_contexts: List[Context] = []
+    incoming_contexts: List[Context] = []  # Default context for all members
     user_prompt: str
     council_models: List[CouncilMemberConfig]  # 2-6 models
-    chairman_model: str  # Model name for synthesis
+    chairman_model: CouncilMemberConfig  # Model config for synthesis
     system_prompt: Optional[str] = None
     temperature: float = 0.7
     reasoning_effort: ReasoningEffort = ReasoningEffort.MEDIUM
+    enable_google_search: bool = False
     merge_strategy: MergeStrategy = MergeStrategy.SMART
     attached_file_ids: List[str] = Field(default_factory=list)
     scopes: List[str] = Field(default_factory=list)
diff --git a/backend/app/services/council.py b/backend/app/services/council.py
index d177f44..ef6fa89 100644
--- a/backend/app/services/council.py
+++ b/backend/app/services/council.py
@@ -21,18 +21,20 @@ async def stage1_collect_responses(
     tools_per_model: Optional[List[Optional[List[Dict[str, Any]]]]] = None,
     openrouter_api_key: Optional[str] = None,
     images: Optional[List[Dict[str, Any]]] = None,
+    contexts_per_model: Optional[List[Optional[Context]]] = None,
 ) -> AsyncGenerator[Dict[str, Any], None]:
     """
     Stage 1: Query all council member models in parallel.
     Yields events as each model completes.
-    Returns final list via stage1_complete event.
+    If contexts_per_model is provided, each member uses its own context (None = shared context).
     """
     async def _query_one(idx: int, config: LLMConfig) -> Dict[str, Any]:
         atts = attachments_per_model[idx] if attachments_per_model else None
         tls = tools_per_model[idx] if tools_per_model else None
+        member_context = (contexts_per_model[idx] if contexts_per_model and idx < len(contexts_per_model) and contexts_per_model[idx] is not None else context)
         try:
             response = await query_model_full(
-                context, user_prompt, config,
+                member_context, user_prompt, config,
                 attachments=atts, tools=tls,
                 openrouter_api_key=openrouter_api_key, images=images,
             )
@@ -255,9 +257,11 @@ async def council_event_stream(
     tools_per_model: Optional[List[Optional[List[Dict[str, Any]]]]] = None,
     openrouter_api_key: Optional[str] = None,
     images: Optional[List[Dict[str, Any]]] = None,
+    contexts_per_model: Optional[List[Optional[Context]]] = None,
 ) -> AsyncGenerator[str, None]:
     """
     Master orchestrator yielding SSE JSON events through the 3-stage council process.
+    If contexts_per_model is provided, each member uses its own context (None entries = shared context).
     """
     # === Stage 1 ===
     yield _sse_event({"type": "stage1_start"})
@@ -269,6 +273,7 @@ async def council_event_stream(
         tools_per_model=tools_per_model,
         openrouter_api_key=openrouter_api_key,
         images=images,
+        contexts_per_model=contexts_per_model,
     ):
         stage1_results.append(result)
         yield _sse_event({