scaleapi
diff --git a/examples/demos/deep_research/README.md b/examples/demos/deep_research/README.md
@@ -0,0 +1,143 @@
+# Deep Research Multi-Agent System
+
+A multi-agent research system built on AgentEx that demonstrates **orchestrator + subagent communication** using Temporal workflows. An orchestrator agent dispatches specialized research subagents (GitHub, Docs, Slack) in parallel, collects their findings, and synthesizes a comprehensive answer.
+
+## Architecture
+
+```
+                    ┌─────────────────────┐
+                    │   Orchestrator      │
+          User ────▶│   (GPT-5.1)         │
+          Query     │   Dispatches &      │
+                    │   Synthesizes       │
+                    └───┬─────┬─────┬─────┘
+                        │     │     │
+              ┌─────────┘     │     └─────────┐
+              ▼               ▼               ▼
+     ┌────────────┐  ┌────────────┐  ┌────────────┐
+     │  GitHub    │  │   Docs     │  │   Slack    │
+     │ Researcher │  │ Researcher │  │ Researcher │
+     │ (GPT-4.1  │  │ (GPT-4.1  │  │ (GPT-4.1  │
+     │  mini)    │  │  mini)    │  │  mini)    │
+     │           │  │           │  │           │
+     │ GitHub MCP│  │ Web Search│  │ Slack MCP │
+     │ Server    │  │ + Fetcher │  │ Server    │
+     └────────────┘  └────────────┘  └────────────┘
+```
+
+## Key Patterns Demonstrated
+
+### 1. Multi-Agent Orchestration via ACP
+The orchestrator creates child tasks on subagents using `adk.acp.create_task()`, sends queries via `EVENT_SEND`, and waits for `research_complete` callback events.
+
+### 2. Shared Task ID for Unified Output
+All subagents write messages to the **orchestrator's task ID** (passed as `source_task_id`), so the user sees all research progress in a single conversation thread.
+
+### 3. Conversation Compaction
+Subagents use a batched `Runner.run()` pattern with conversation compaction between batches to stay within Temporal's ~2MB payload limit during long research sessions.
+
+### 4. MCP Server Integration
+GitHub and Slack subagents use MCP (Model Context Protocol) servers via `StatelessMCPServerProvider` for tool access.
+
+## Agents
+
+| Agent | Port | Model | Tools |
+|-------|------|-------|-------|
+| Orchestrator | 8010 | gpt-5.1 | dispatch_github, dispatch_docs, dispatch_slack |
+| GitHub Researcher | 8011 | gpt-4.1-mini | GitHub MCP (search_code, etc.) |
+| Docs Researcher | 8012 | gpt-4.1-mini | web_search (Tavily), fetch_docs_page |
+| Slack Researcher | 8013 | gpt-4.1-mini | Slack MCP (search_messages, etc.) |
+
+## Prerequisites
+
+- [AgentEx CLI](https://agentex.sgp.scale.com/docs/) installed
+- OpenAI API key
+- GitHub Personal Access Token (for GitHub researcher)
+- Tavily API key (for Docs researcher) - get one at https://tavily.com
+- Slack Bot Token (for Slack researcher)
+
+## Setup
+
+### 1. Environment Variables
+
+Create a `.env` file in each agent directory with the required keys:
+
+**orchestrator/.env:**
+```
+OPENAI_API_KEY=your-openai-key
+```
+
+**github_researcher/.env:**
+```
+OPENAI_API_KEY=your-openai-key
+GITHUB_PERSONAL_ACCESS_TOKEN=your-github-token
+```
+
+**docs_researcher/.env:**
+```
+OPENAI_API_KEY=your-openai-key
+TAVILY_API_KEY=your-tavily-key
+```
+
+**slack_researcher/.env:**
+```
+OPENAI_API_KEY=your-openai-key
+SLACK_BOT_TOKEN=your-slack-bot-token
+SLACK_TEAM_ID=your-slack-team-id
+```
+
+### 2. Run All Agents
+
+Start each agent in a separate terminal:
+
+```bash
+# Terminal 1 - Orchestrator
+cd orchestrator
+agentex agents run --manifest manifest.yaml
+
+# Terminal 2 - GitHub Researcher
+cd github_researcher
+agentex agents run --manifest manifest.yaml
+
+# Terminal 3 - Docs Researcher
+cd docs_researcher
+agentex agents run --manifest manifest.yaml
+
+# Terminal 4 - Slack Researcher
+cd slack_researcher
+agentex agents run --manifest manifest.yaml
+```
+
+### 3. Test
+
+Open the AgentEx UI and send a research question to the orchestrator agent. You should see:
+1. The orchestrator dispatching queries to subagents
+2. Each subagent streaming its research progress to the same conversation
+3. The orchestrator synthesizing all findings into a final answer
+
+## Customization
+
+### Using Different Research Sources
+
+You can adapt the subagents to search different sources:
+- Replace the GitHub MCP server with any other MCP server
+- Replace Tavily with your preferred search API
+- Replace the Slack MCP with any communication platform's MCP
+- Update the system prompts to match your target repositories, docs, and channels
+
+### Adding More Subagents
+
+To add a new research subagent:
+1. Copy one of the existing subagent directories
+2. Update the manifest.yaml with a new agent name and port
+3. Modify the workflow.py system prompt and tools
+4. Add a new dispatch tool in the orchestrator's workflow.py
+
+## How Shared Task ID Works
+
+The key pattern that makes all agents write to the same conversation:
+
+1. **Orchestrator** passes its `task_id` as `source_task_id` when creating child tasks
+2. **Subagents** extract `parent_task_id = params.task.params.get("source_task_id")`
+3. **Subagents** use `message_task_id = parent_task_id or params.task.id` for all `adk.messages.create()` calls and `TemporalStreamingHooks`
+4. This means all messages and streamed LLM output appear in the orchestrator's task conversation
diff --git a/examples/demos/deep_research/docs_researcher/.dockerignore b/examples/demos/deep_research/docs_researcher/.dockerignore
@@ -0,0 +1,43 @@
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# Environments
+.env**
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# IDE
+.idea/
+.vscode/
+*.swp
+*.swo
+
+# Git
+.git
+.gitignore
+
+# Misc
+.DS_Store
diff --git a/examples/demos/deep_research/docs_researcher/Dockerfile b/examples/demos/deep_research/docs_researcher/Dockerfile
@@ -0,0 +1,40 @@
+# syntax=docker/dockerfile:1.3
+FROM python:3.12-slim
+COPY --from=ghcr.io/astral-sh/uv:0.6.4 /uv /uvx /bin/
+
+# Install system dependencies
+RUN apt-get update && apt-get install -y \
+    htop \
+    vim \
+    curl \
+    tar \
+    python3-dev \
+    build-essential \
+    gcc \
+    cmake \
+    netcat-openbsd \
+    && apt-get clean \
+    && rm -rf /var/lib/apt/lists/**
+
+# Install tctl (Temporal CLI)
+RUN curl -L https://github.com/temporalio/tctl/releases/download/v1.18.1/tctl_1.18.1_linux_arm64.tar.gz -o /tmp/tctl.tar.gz && \
+    tar -xzf /tmp/tctl.tar.gz -C /usr/local/bin && \
+    chmod +x /usr/local/bin/tctl && \
+    rm /tmp/tctl.tar.gz
+
+RUN uv pip install --system --upgrade pip setuptools wheel
+
+ENV UV_HTTP_TIMEOUT=1000
+
+COPY deep_research/docs_researcher/pyproject.toml /app/docs_researcher/pyproject.toml
+
+WORKDIR /app/docs_researcher
+
+COPY deep_research/docs_researcher/project /app/docs_researcher/project
+
+RUN uv pip install --system .
+
+ENV PYTHONPATH=/app
+ENV AGENT_NAME=deep-research-docs
+
+CMD ["uvicorn", "project.acp:acp", "--host", "0.0.0.0", "--port", "8000"]
diff --git a/examples/demos/deep_research/docs_researcher/environments.yaml b/examples/demos/deep_research/docs_researcher/environments.yaml
@@ -0,0 +1,27 @@
+# Agent Environment Configuration
+schema_version: "v1"
+environments:
+  dev:
+    auth:
+      principal:
+        user_id: # TODO: Fill in
+        account_id: # TODO: Fill in
+    helm_overrides:
+      replicaCount: 1
+      resources:
+        requests:
+          cpu: "500m"
+          memory: "1Gi"
+        limits:
+          cpu: "1000m"
+          memory: "2Gi"
+      temporal-worker:
+        enabled: true
+        replicaCount: 1
+        resources:
+          requests:
+            cpu: "500m"
+            memory: "1Gi"
+          limits:
+            cpu: "1000m"
+            memory: "2Gi"
diff --git a/examples/demos/deep_research/docs_researcher/manifest.yaml b/examples/demos/deep_research/docs_researcher/manifest.yaml
@@ -0,0 +1,59 @@
+# Agent Manifest Configuration
+
+build:
+  context:
+    root: ../../
+    include_paths:
+      - deep_research/docs_researcher
+    dockerfile: deep_research/docs_researcher/Dockerfile
+    dockerignore: deep_research/docs_researcher/.dockerignore
+
+local_development:
+  agent:
+    port: 8012
+    host_address: host.docker.internal
+
+  paths:
+    acp: project/acp.py
+    worker: project/run_worker.py
+
+agent:
+  acp_type: async
+  name: deep-research-docs
+  description: Searches documentation and the web for relevant guides and references
+
+  temporal:
+    enabled: true
+    workflows:
+      - name: deep-research-docs
+        queue_name: deep_research_docs_queue
+
+  credentials:
+    - env_var_name: REDIS_URL
+      secret_name: redis-url-secret
+      secret_key: url
+
+  env:
+    OPENAI_ORG_ID: ""
+
+deployment:
+  image:
+    repository: ""
+    tag: "latest"
+
+  imagePullSecrets:
+    - name: my-registry-secret
+
+  global:
+    agent:
+      name: "deep-research-docs"
+      description: "Searches documentation and the web for relevant guides and references"
+
+    replicaCount: 1
+    resources:
+      requests:
+        cpu: "500m"
+        memory: "1Gi"
+      limits:
+        cpu: "1000m"
+        memory: "2Gi"
diff --git a/examples/demos/deep_research/docs_researcher/project/__init__.py b/examples/demos/deep_research/docs_researcher/project/__init__.py
diff --git a/examples/demos/deep_research/docs_researcher/project/acp.py b/examples/demos/deep_research/docs_researcher/project/acp.py
@@ -0,0 +1,29 @@
+import os
+
+from temporalio.contrib.openai_agents import OpenAIAgentsPlugin
+
+from agentex.lib.core.temporal.plugins.openai_agents.models.temporal_streaming_model import (
+    TemporalStreamingModelProvider,
+)
+from agentex.lib.core.temporal.plugins.openai_agents.interceptors.context_interceptor import ContextInterceptor
+from agentex.lib.sdk.fastacp.fastacp import FastACP
+from agentex.lib.types.fastacp import TemporalACPConfig
+
+context_interceptor = ContextInterceptor()
+streaming_model_provider = TemporalStreamingModelProvider()
+
+# Create the ACP server
+acp = FastACP.create(
+    acp_type="async",
+    config=TemporalACPConfig(
+        type="temporal",
+        temporal_address=os.getenv("TEMPORAL_ADDRESS", "localhost:7233"),
+        plugins=[OpenAIAgentsPlugin(model_provider=streaming_model_provider)],
+        interceptors=[context_interceptor],
+    ),
+)
+
+if __name__ == "__main__":
+    import uvicorn
+
+    uvicorn.run(acp, host="0.0.0.0", port=8000)