fix(chat+langgraph): streaming complex-content + GFM markdown + model picker (chat 0.0.18, langgraph 0.0.10)

cockpit/chat/generative-ui/python/src/graph.py

@@ -19,7 +19,7 @@
 
 _PROMPT = (Path(__file__).parent.parent / "prompts" / "dashboard.md").read_text()
 
-_llm = ChatOpenAI(model="gpt-4o-mini", temperature=0, streaming=True)
+_llm = ChatOpenAI(model="gpt-5-mini", temperature=0, streaming=True)
 _llm_with_tools = _llm.bind_tools(ALL_TOOLS)
 
 

cockpit/deep-agents/filesystem/python/docs/guide.md

@@ -139,7 +139,7 @@ def write_file(path: str, content: str) -> str:
     return f"Successfully wrote {len(content)} bytes to {path}"
 
 # Bind tools to the LLM
-llm = ChatOpenAI(model="gpt-4o-mini").bind_tools([read_file, write_file])
+llm = ChatOpenAI(model="gpt-5-mini").bind_tools([read_file, write_file])
 ```
 
 The agent node invokes the LLM, which may emit tool calls. A conditional edge routes to the `ToolNode` when tool calls are present, then loops back to the agent. The frontend sees each tool call in `stream.messages()`.

cockpit/deep-agents/filesystem/python/src/graph.py

@@ -35,7 +35,7 @@ class FilesystemState(TypedDict):
 
 def build_filesystem_graph():
     tools = [read_file, write_file]
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True).bind_tools(tools)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True).bind_tools(tools)
 
     async def agent(state: FilesystemState) -> dict:
         """Run the agent — may emit tool calls."""

cockpit/deep-agents/memory/python/src/graph.py

@@ -24,7 +24,7 @@ class MemoryState(TypedDict):
 
 
 def build_memory_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state: MemoryState) -> dict:
         """Generate a response using remembered facts in the system prompt."""

cockpit/deep-agents/planning/python/src/graph.py

@@ -27,7 +27,7 @@ class PlanningState(TypedDict):
 
 
 def build_planning_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def create_plan(state: PlanningState) -> dict:
         """Decompose the task into ordered steps."""

cockpit/deep-agents/sandboxes/python/docs/guide.md

@@ -153,7 +153,7 @@ def run_code(code: str) -> str:
         "exit_status": 0,
     })
 
-llm = ChatOpenAI(model="gpt-4o-mini").bind_tools([run_code])
+llm = ChatOpenAI(model="gpt-5-mini").bind_tools([run_code])
 tool_node = ToolNode([run_code])
 ```
 

cockpit/deep-agents/sandboxes/python/src/graph.py

@@ -71,7 +71,7 @@ def run_code(code: str) -> str:
 
 def build_sandboxes_graph():
     tools = [run_code]
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True).bind_tools(tools)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True).bind_tools(tools)
     tool_node = ToolNode(tools)
 
     async def agent(state: SandboxesState) -> dict:

cockpit/deep-agents/skills/python/docs/guide.md

@@ -150,7 +150,7 @@ def summarize(text: str) -> str:
     return sentences[0] + "." if sentences else "No content."
 
 # Bind all tools to the LLM
-llm = ChatOpenAI(model="gpt-4o-mini").bind_tools([calculator, word_count, summarize])
+llm = ChatOpenAI(model="gpt-5-mini").bind_tools([calculator, word_count, summarize])
 ```
 
 The agent selects which skill to call based on the user's request. `ToolNode` dispatches the call and returns the result as a `ToolMessage`.

cockpit/deep-agents/skills/python/src/graph.py

@@ -65,7 +65,7 @@ def summarize(text: str) -> str:
 
 def build_skills_graph():
     tools = [calculator, word_count, summarize]
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True).bind_tools(tools)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True).bind_tools(tools)
     tool_node = ToolNode(tools)
 
     async def agent(state: SkillsState) -> dict:

cockpit/deep-agents/subagents/python/src/graph.py

@@ -23,12 +23,12 @@ class SubagentsState(TypedDict):
 
 
 def build_subagents_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     @tool
     async def research_agent(topic: str) -> str:
         """Spawn a research subagent to gather information on a topic."""
-        research_llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+        research_llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
         response = await research_llm.ainvoke([
             SystemMessage(content="You are a research specialist. Provide concise, factual information."),
             {"role": "human", "content": f"Research this topic and provide key facts: {topic}"},
@@ -38,7 +38,7 @@ async def research_agent(topic: str) -> str:
     @tool
     async def analysis_agent(content: str) -> str:
         """Spawn an analysis subagent to analyze and synthesize information."""
-        analysis_llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+        analysis_llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
         response = await analysis_llm.ainvoke([
             SystemMessage(content="You are an analysis specialist. Identify patterns, draw insights, and synthesize information clearly."),
             {"role": "human", "content": f"Analyze this content and provide key insights: {content}"},
@@ -48,7 +48,7 @@ async def analysis_agent(content: str) -> str:
     @tool
     async def summary_agent(findings: str) -> str:
         """Spawn a summary subagent to produce a final coherent response."""
-        summary_llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+        summary_llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
         response = await summary_llm.ainvoke([
             SystemMessage(content="You are a summarization specialist. Produce clear, well-structured summaries."),
             {"role": "human", "content": f"Summarize these findings into a concise final answer: {findings}"},

cockpit/langgraph/deployment-runtime/python/src/graph.py

@@ -28,7 +28,7 @@ def build_deployment_runtime_graph():
     LangGraph Cloud. The assistantId in the Angular component must match
     the graph key in langgraph.json.
     """
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state: MessagesState) -> dict:
         """Generate a response using the full message history."""

cockpit/langgraph/durable-execution/python/docs/guide.md

@@ -124,7 +124,7 @@ class DurableState(TypedDict):
 checkpointer = MemorySaver()
 
 def build_durable_execution_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def analyze(state): ...   # Node 1
     async def plan(state):   ...   # Node 2

cockpit/langgraph/durable-execution/python/src/graph.py

@@ -32,7 +32,7 @@ def build_durable_execution_graph():
     and checkpoints state after each one. This means any node failure
     only requires replaying from the previous checkpoint, not the start.
     """
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
     system_prompt = (PROMPTS_DIR / "durable-execution.md").read_text()
 
     async def analyze(state: DurableState) -> dict:

cockpit/langgraph/interrupts/python/docs/guide.md

@@ -111,7 +111,7 @@ from langgraph.types import interrupt
 checkpointer = MemorySaver()
 
 def build_interrupts_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state: MessagesState) -> dict:
         response = await llm.ainvoke(state["messages"])

cockpit/langgraph/memory/python/docs/guide.md

@@ -119,7 +119,7 @@ class MemoryState(TypedDict):
 checkpointer = MemorySaver()
 
 def build_memory_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state: MemoryState) -> dict:
         memory = state.get("memory", {})

cockpit/langgraph/memory/python/src/graph.py

@@ -37,8 +37,8 @@ def build_memory_graph():
     This ensures every response is followed by a memory extraction pass,
     keeping the agent's knowledge up to date without blocking the reply.
     """
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
-    extractor_llm = ChatOpenAI(model="gpt-4o-mini", streaming=False)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
+    extractor_llm = ChatOpenAI(model="gpt-5-mini", streaming=False)
 
     async def generate(state: MemoryState) -> dict:
         """Generate a response using current messages and known memory."""

cockpit/langgraph/persistence/python/docs/guide.md

@@ -121,7 +121,7 @@ from langgraph.checkpoint.memory import MemorySaver
 checkpointer = MemorySaver()
 
 def build_persistence_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state: MessagesState) -> dict:
         response = await llm.ainvoke(state["messages"])

cockpit/langgraph/streaming/python/docs/guide.md

@@ -95,7 +95,7 @@ from langgraph.graph import StateGraph, END
 from langchain_openai import ChatOpenAI
 
 def build_streaming_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state):
         response = await llm.ainvoke(state["messages"])

cockpit/langgraph/streaming/python/src/dashboard_graph.py

@@ -19,7 +19,7 @@
 
 _PROMPT = (Path(__file__).parent.parent / "prompts" / "dashboard.md").read_text()
 
-_llm = ChatOpenAI(model="gpt-4o-mini", temperature=0, streaming=True)
+_llm = ChatOpenAI(model="gpt-5-mini", temperature=0, streaming=True)
 _llm_with_tools = _llm.bind_tools(ALL_TOOLS)
 
 

cockpit/langgraph/subgraphs/python/docs/guide.md

@@ -89,7 +89,7 @@ The backend uses a parent orchestrator that delegates to a compiled child subgra
 from langgraph.graph import StateGraph, MessagesState, END
 from langchain_openai import ChatOpenAI
 
-llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
 # Child: research subgraph
 async def research_node(state: MessagesState) -> dict:

cockpit/langgraph/subgraphs/python/src/graph.py

@@ -22,7 +22,7 @@ def build_subgraphs_graph():
     The parent orchestrator decides when to delegate to the research subgraph.
     The research subgraph runs independently and returns its results to the parent.
     """
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     # ── Child: research subgraph ──────────────────────────────────────────────
 

cockpit/langgraph/time-travel/python/docs/guide.md

@@ -112,7 +112,7 @@ from langgraph.checkpoint.memory import MemorySaver
 checkpointer = MemorySaver()
 
 def build_time_travel_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state: MessagesState) -> dict:
         response = await llm.ainvoke(state["messages"])

cockpit/langgraph/time-travel/python/src/graph.py

@@ -25,7 +25,7 @@ def build_time_travel_graph():
     producing a history of ThreadState objects that the client can replay or
     branch from using checkpoint IDs.
     """
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state: MessagesState) -> dict:
         """Generate a response, checkpointed for time travel."""

docs/superpowers/plans/2026-03-19-langsmith-deployment.md

@@ -20,7 +20,7 @@ Website (Vercel)
        └─ NEXT_PUBLIC_LANGGRAPH_URL
             └─ LangGraph Cloud (LangSmith)
                  └─ chat_agent graph (Python)
-                      └─ ChatOpenAI (gpt-4o-mini)
+                      └─ ChatOpenAI (gpt-5-mini)
 ```
 
 ---
@@ -141,7 +141,7 @@ From the LangSmith dashboard → Deployments → your deployment → Environment
 | Variable | Value |
 |---|---|
 | `OPENAI_API_KEY` | Your OpenAI key |
-| `OPENAI_MODEL` | `gpt-4o-mini` (default) or preferred model |
+| `OPENAI_MODEL` | `gpt-5-mini` (default) or preferred model |
 | `LANGSMITH_TRACING` | `true` (optional, for production tracing) |
 | `LANGSMITH_PROJECT` | `angular-example` |
 

docs/superpowers/plans/2026-04-03-cockpit-example-harness.md

@@ -1077,7 +1077,7 @@ def build_streaming_graph() -> StateGraph:
     Returns:
         A compiled StateGraph ready for invocation
     """
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state: StreamingState) -> dict:
         """

docs/superpowers/plans/2026-04-03-docs-mode-redesign.md

@@ -779,7 +779,7 @@ from langgraph.graph import StateGraph, END
 from langchain_openai import ChatOpenAI
 
 def build_streaming_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state):
         response = await llm.ainvoke(state["messages"])

docs/superpowers/plans/2026-04-03-narrative-docs.md

@@ -514,7 +514,7 @@ from langgraph.graph import StateGraph, END
 from langchain_openai import ChatOpenAI
 
 def build_streaming_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state):
         response = await llm.ainvoke(state["messages"])

docs/superpowers/plans/2026-04-08-generative-ui-spike.md

@@ -176,7 +176,7 @@ PROMPTS_DIR = Path(__file__).parent.parent / "prompts"
 
 
 def build_generative_ui_graph():
-    llm = ChatOpenAI(model="gpt-4o-mini", streaming=True)
+    llm = ChatOpenAI(model="gpt-5-mini", streaming=True)
 
     async def generate(state: MessagesState) -> dict:
         system_prompt = (PROMPTS_DIR / "generative-ui.md").read_text()