Python: DevUI Fix Serialization, Timestamp and Other Issues (#1584)

* refactor(devui): adopt standard OpenAI lifecycle events for agents and workflows

- Replace custom workflow events with OpenAI Responses API standard lifecycle events
- Add AgentStartedEvent, AgentCompletedEvent, AgentFailedEvent for clean separation
- Implement ExecutorActionItem for workflow executor tracking
- Convert informational events to trace events to reduce noise
- Update README mapper table with comprehensive event mappings
- Maintain full backward compatibility with legacy events

* fix(devui): resolve timestamp overwriting and Content serialization errors

- Fix tool call timestamps being overwritten on each render (#1483)
- Add recursive Content serialization to handle ChatMessage and nested objects (#1548)
- Implement proper MCP tool cleanup on server shutdown
- Add timestamp field to function_result.complete events
- Enhance credential and client resource cleanup

Fixes #1483, #1548
Partial improvements for #1476
This commit is contained in:
Victor Dibia
2025-10-23 11:19:20 -07:00
committed by GitHub
Unverified
parent 064ee8afbe
commit 6b66a34609
21 changed files with 1859 additions and 682 deletions
+55 -20
View File
@@ -49,6 +49,19 @@ devui ./agents --port 8080
When DevUI starts with no discovered entities, it displays a **sample entity gallery** with curated examples from the Agent Framework repository. You can download these samples, review them, and run them locally to get started quickly.
## Using MCP Tools
**Important:** Don't use `async with` context managers when creating agents with MCP tools for DevUI - connections will close before execution.
```python
# ✅ Correct - DevUI handles cleanup automatically
mcp_tool = MCPStreamableHTTPTool(url="http://localhost:8011/mcp", chat_client=chat_client)
agent = ChatAgent(tools=mcp_tool)
serve(entities=[agent])
```
MCP tools use lazy initialization and connect automatically on first use. DevUI attempts to clean up connections on shutdown
## Directory Structure
For your agents to be discovered by the DevUI, they must be organized in a directory structure like below. Each agent/workflow must have an `__init__.py` that exports the required variable (`agent` or `workflow`).
@@ -157,42 +170,62 @@ Options:
Given that DevUI offers an OpenAI Responses API, it internally maps messages and events from Agent Framework to OpenAI Responses API events (in `_mapper.py`). For transparency, this mapping is shown below:
| Agent Framework Content | OpenAI Event/Type | Status |
| ------------------------------- | ---------------------------------------- | -------- |
| `TextContent` | `response.output_text.delta` | Standard |
| `TextReasoningContent` | `response.reasoning_text.delta` | Standard |
| `FunctionCallContent` (initial) | `response.output_item.added` | Standard |
| `FunctionCallContent` (args) | `response.function_call_arguments.delta` | Standard |
| `FunctionResultContent` | `response.function_result.complete` | DevUI |
| `FunctionApprovalRequestContent`| `response.function_approval.requested` | DevUI |
| `FunctionApprovalResponseContent`| `response.function_approval.responded` | DevUI |
| `ErrorContent` | `error` | Standard |
| `UsageContent` | Final `Response.usage` field (not streamed) | Standard |
| `WorkflowEvent` | `response.workflow_event.complete` | DevUI |
| `DataContent` | `response.trace.complete` | DevUI |
| `UriContent` | `response.trace.complete` | DevUI |
| `HostedFileContent` | `response.trace.complete` | DevUI |
| `HostedVectorStoreContent` | `response.trace.complete` | DevUI |
| OpenAI Event/Type | Agent Framework Content | Status |
| ------------------------------------------------------------ | --------------------------------- | -------- |
| | **Lifecycle Events** | |
| `response.created` + `response.in_progress` | `AgentStartedEvent` | OpenAI |
| `response.completed` | `AgentCompletedEvent` | OpenAI |
| `response.failed` | `AgentFailedEvent` | OpenAI |
| `response.created` + `response.in_progress` | `WorkflowStartedEvent` | OpenAI |
| `response.completed` | `WorkflowCompletedEvent` | OpenAI |
| `response.failed` | `WorkflowFailedEvent` | OpenAI |
| | **Content Types** | |
| `response.content_part.added` + `response.output_text.delta` | `TextContent` | OpenAI |
| `response.reasoning_text.delta` | `TextReasoningContent` | OpenAI |
| `response.output_item.added` | `FunctionCallContent` (initial) | OpenAI |
| `response.function_call_arguments.delta` | `FunctionCallContent` (args) | OpenAI |
| `response.function_result.complete` | `FunctionResultContent` | DevUI |
| `response.function_approval.requested` | `FunctionApprovalRequestContent` | DevUI |
| `response.function_approval.responded` | `FunctionApprovalResponseContent` | DevUI |
| `error` | `ErrorContent` | OpenAI |
| Final `Response.usage` field (not streamed) | `UsageContent` | OpenAI |
| | **Workflow Events** | |
| `response.output_item.added` (ExecutorActionItem)* | `ExecutorInvokedEvent` | OpenAI |
| `response.output_item.done` (ExecutorActionItem)* | `ExecutorCompletedEvent` | OpenAI |
| `response.output_item.done` (ExecutorActionItem with error)* | `ExecutorFailedEvent` | OpenAI |
| `response.workflow_event.complete` | `WorkflowEvent` (other) | DevUI |
| `response.trace.complete` | `WorkflowStatusEvent` | DevUI |
| `response.trace.complete` | `WorkflowWarningEvent` | DevUI |
| | **Trace Content** | |
| `response.trace.complete` | `DataContent` | DevUI |
| `response.trace.complete` | `UriContent` | DevUI |
| `response.trace.complete` | `HostedFileContent` | DevUI |
| `response.trace.complete` | `HostedVectorStoreContent` | DevUI |
- **Standard** = OpenAI Responses API spec
- **DevUI** = Custom extensions for Agent Framework features (workflows, traces, function approvals)
\*Uses standard OpenAI event structure but carries DevUI-specific `ExecutorActionItem` payload
- **OpenAI** = Standard OpenAI Responses API event types
- **DevUI** = Custom event types specific to Agent Framework (e.g., workflows, traces, function approvals)
### OpenAI Responses API Compliance
DevUI follows the OpenAI Responses API specification for maximum compatibility:
**Standard OpenAI Types Used:**
**OpenAI Standard Event Types Used:**
- `ResponseOutputItemAddedEvent` - Output item notifications (function calls and results)
- `ResponseOutputItemDoneEvent` - Output item completion notifications
- `Response.usage` - Token usage (in final response, not streamed)
- All standard text, reasoning, and function call events
**Custom DevUI Extensions:**
- `response.function_approval.requested` - Function approval requests (for interactive approval workflows)
- `response.function_approval.responded` - Function approval responses (user approval/rejection)
- `response.workflow_event.complete` - Agent Framework workflow events
- `response.trace.complete` - Execution traces and internal content (DataContent, UriContent, hosted files/stores)
These custom extensions are clearly namespaced and can be safely ignored by standard OpenAI clients.
These custom extensions are clearly namespaced and can be safely ignored by standard OpenAI clients. Note that DevUI also uses standard OpenAI events with custom payloads (e.g., `ExecutorActionItem` within `response.output_item.added`).
### Entity Management
@@ -224,12 +257,14 @@ These custom extensions are clearly namespaced and can be safely ignored by stan
DevUI is designed as a **sample application for local development** and should not be exposed to untrusted networks or used in production environments.
**Security features:**
- Only loads entities from local directories or in-memory registration
- No remote code execution capabilities
- Binds to localhost (127.0.0.1) by default
- All samples must be manually downloaded and reviewed before running
**Best practices:**
- Never expose DevUI to the internet
- Review all agent/workflow code before running
- Only load entities from trusted sources