Remove max_concurrency, not supported (yet) when we .map() in the beta Graph API

docs/agents.md

@@ -59,7 +59,7 @@ stateDiagram-v2
 Key nodes:
 
 - **Plan**: Decomposes the question into focused sub-questions
-- **Search (batched)**: Answers sub-questions in parallel batches (respects max_concurrency)
+- **Search (parallel)**: Answers all sub-questions in parallel
 - **Decision**: Evaluates if we have sufficient information or need another iteration
 - **Synthesize**: Generates the final comprehensive answer
 
@@ -69,7 +69,7 @@ Key differences from Research:
 - **Direct answers**: Returns just the answer (not a full research report)
 - **Question-focused**: Optimized for answering specific questions, not open-ended research
 - **Supports citations**: Can include inline source citations like `[document.md]`
-- **Configurable iterations**: Control max_iterations (default: 2) and max_concurrency (default: 3)
+- **Configurable iterations**: Control max_iterations (default: 2)
 
 CLI usage:
 
@@ -99,8 +99,7 @@ async with HaikuRAG(path_to_db) as client:
     state = DeepQAState(
         context=context,
         max_sub_questions=3,
-        max_iterations=2,
+        max_iterations=2
-        max_concurrency=3
     )
     deps = DeepQADeps(client=client)
 
@@ -176,7 +175,6 @@ async with HaikuRAG(path_to_db) as client:
         context=ResearchContext(original_question=question),
         max_iterations=2,
         confidence_threshold=0.8,
-        max_concurrency=2,
     )
     deps = ResearchDeps(client=client)
 
@@ -211,7 +209,6 @@ async with HaikuRAG(path_to_db) as client:
         context=ResearchContext(original_question=question),
         max_iterations=2,
         confidence_threshold=0.8,
-        max_concurrency=2,
     )
     deps = ResearchDeps(client=client)
 

haiku_rag_slim/haiku/rag/app.py

@@ -242,7 +242,6 @@ class HaikuRAGApp:
         question: str,
         max_iterations: int = 3,
         confidence_threshold: float = 0.8,
-        max_concurrency: int = 1,
         verbose: bool = False,
     ):
         """Run research via the pydantic-graph pipeline (default)."""
@@ -262,7 +261,6 @@ class HaikuRAGApp:
                     context=context,
                     max_iterations=max_iterations,
                     confidence_threshold=confidence_threshold,
-                    max_concurrency=max_concurrency,
                 )
                 deps = ResearchDeps(
                     client=client, console=self.console if verbose else None

haiku_rag_slim/haiku/rag/cli.py

@@ -301,11 +301,6 @@ def research(
         "--confidence-threshold",
         help="Minimum confidence (0-1) to stop",
     ),
-    max_concurrency: int = typer.Option(
-        1,
-        "--max-concurrency",
-        help="Max concurrent searches per iteration (planned)",
-    ),
     db: Path = typer.Option(
         Config.storage.data_dir / "haiku.rag.lancedb",
         "--db",
@@ -325,7 +320,6 @@ def research(
             question=question,
             max_iterations=max_iterations,
             confidence_threshold=confidence_threshold,
-            max_concurrency=max_concurrency,
             verbose=verbose,
         )
     )

haiku_rag_slim/haiku/rag/mcp.py

@@ -219,7 +219,6 @@ def create_mcp_server(db_path: Path) -> FastMCP:
         question: str,
         max_iterations: int = 3,
         confidence_threshold: float = 0.8,
-        max_concurrency: int = 1,
     ) -> ResearchReport | None:
         """Run multi-agent research to investigate a complex question.
 
@@ -230,7 +229,6 @@ def create_mcp_server(db_path: Path) -> FastMCP:
             question: The research question to investigate.
             max_iterations: Maximum search/analyze iterations (default: 3).
             confidence_threshold: Minimum confidence score (0-1) to stop early (default: 0.8).
-            max_concurrency: Maximum concurrent searches per iteration (default: 1).
 
         Returns:
             A research report with findings, or None if an error occurred.
@@ -249,7 +247,6 @@ def create_mcp_server(db_path: Path) -> FastMCP:
                     context=ResearchContext(original_question=question),
                     max_iterations=max_iterations,
                     confidence_threshold=confidence_threshold,
-                    max_concurrency=max_concurrency,
                 )
                 deps = ResearchDeps(client=rag)
 

haiku_rag_slim/haiku/rag/qa/deep/graph.py

@@ -149,13 +149,16 @@ def build_deep_qa_graph(
     async def get_batch(
         ctx: StepContext[DeepQAState, DeepQADeps, None | bool],
     ) -> list[str] | None:
-        """Get next batch of questions from state."""
+        """Get all remaining questions for this iteration."""
         state = ctx.state
-        take = max(1, state.max_concurrency)
+
-        batch: list[str] = []
+        if not state.context.sub_questions:
-        while state.context.sub_questions and len(batch) < take:
+            return None
-            batch.append(state.context.sub_questions.pop(0))
+
-        return batch if batch else None
+        # Take ALL remaining questions - max_concurrency controls parallel execution within .map()
+        batch = list(state.context.sub_questions)
+        state.context.sub_questions.clear()
+        return batch
 
     @g.step
     async def decide(

haiku_rag_slim/haiku/rag/qa/deep/state.py

@@ -21,5 +21,4 @@ class DeepQAState:
     context: DeepQAContext
     max_sub_questions: int = 3
     max_iterations: int = 2
-    max_concurrency: int = 1
     iterations: int = 0

haiku_rag_slim/haiku/rag/research/graph.py

@@ -159,13 +159,16 @@ def build_research_graph(
     async def get_batch(
         ctx: StepContext[ResearchState, ResearchDeps, None | bool],
     ) -> list[str] | None:
-        """Get next batch of questions from state."""
+        """Get all remaining questions for this iteration."""
         state = ctx.state
-        take = max(1, state.max_concurrency)
+
-        batch: list[str] = []
+        if not state.context.sub_questions:
-        while state.context.sub_questions and len(batch) < take:
+            return None
-            batch.append(state.context.sub_questions.pop(0))
+
-        return batch if batch else None
+        # Take ALL remaining questions and process them in parallel
+        batch = list(state.context.sub_questions)
+        state.context.sub_questions.clear()
+        return batch
 
     @g.step
     async def analyze_insights(

haiku_rag_slim/haiku/rag/research/state.py

@@ -26,7 +26,6 @@ class ResearchState:
     context: ResearchContext
     iterations: int = 0
     max_iterations: int = 3
-    max_concurrency: int = 1
     confidence_threshold: float = 0.8
     last_eval: EvaluationResult | None = None
     last_analysis: InsightAnalysis | None = None

haiku_rag_slim/haiku/rag/research/stream.py

@@ -15,7 +15,6 @@ class ResearchStateSnapshot:
     sub_questions: list[str]
     iterations: int
     max_iterations: int
-    max_concurrency: int
     confidence_threshold: float
     pending_sub_questions: int
     answered_questions: int
@@ -38,7 +37,6 @@ class ResearchStateSnapshot:
             sub_questions=list(context.sub_questions),
             iterations=state.iterations,
             max_iterations=state.max_iterations,
-            max_concurrency=state.max_concurrency,
             confidence_threshold=state.confidence_threshold,
             pending_sub_questions=len(context.sub_questions),
             answered_questions=len(context.qa_responses),