Skip to content

feat(frontend): agent generation in the playground + observability - #4810

Merged
mmabrouk merged 6 commits into
big-agentsfrom
fe-feat/agent-playground-generation
Jun 24, 2026
Merged

mmabrouk merged 6 commits into
big-agentsfrom
fe-feat/agent-playground-generation

Conversation

@ardaerzin

@ardaerzin ardaerzin commented Jun 23, 2026 •

Copy link
Copy Markdown
Contributor

What

Frontend layer for agent generation in the playground + the agent observability fixes.
Stacks on the agent chat slice (#4780): fe-feat/agent-chat-ui-slice → feat/agent-playground-ui (#4775) → big-agents (#4791) → main.

FE-only diff (54 files, all web/).

Playground

  • AgentChatPanel as the third generation arm; buildAgentRequest + per-entity isAgentMode; multi-session tabs, attachments, conversation rewind; streaming over the v6 /messages contract (SSE).
  • Message rendering: markdown (tables, headings, code highlight, LaTeX), collapsible thinking block, tool cards, streamed usage/cost in the hover toolbar, failed-run error surfaced from the trace.
  • Session history picker; "open a session from observability" deep link.

Observability

  • Sessions table: renders cell data (shares the page store via context), real height (rows were clipped), session-id truncation.
  • Agent reply shown in Last output, the session-drawer conversation, and Traces → Outputs — read from the nested agent span instead of the root span's streamed-generator repr.
  • "Open in Playground" replays an agent run from its root span (→ agent chat, not a bare completion).

Agent config controls

Taken verbatim from #4775 (AgentConfigControl, McpServerItemControl, SchemaPropertyRenderer, SchemaControls/index.ts, CreateApp*, iconHelpers, appUtils) — byte-identical, so they drop out of the diff once the stack is wired. The only delta is a 2-line gateway-tools wiring on AgentConfigControl (passes gatewayTools to the tool picker so Composio tools show), on top of #4775's version.

Not in this PR

  • Backend — session-id-on-trace is covered by fix/agent-session-id-tracing. (Two optional small SDK fixes — generator-repr guard, v6 finish-part — exist on the working branch if wanted.)

Parked follow-up

"Open in Playground" from an agent trace opens an empty agent config — needs a committed agent revision with refs on its spans (or a builtin agent schema). Tracked in docs/design/agent-workflows/.

mmabrouk added 5 commits June 19, 2026 21:37
Ports the AgentChatSlice frontend (ai-sdk useChat over the agent /messages
stream) and the RAG_QA_chatbot example onto the current functional agent
branches. See the PR description for the note to Arda.
The agent-chat slice transport relied on the server defaulting to a
stream. The real /messages endpoint only streams when the client sends
Accept: text/event-stream; otherwise it returns a batch JSON body, which
useChat cannot render. Send the SSE Accept header so streamed assistant
replies show up.
@vercel

vercel Bot commented Jun 23, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated (UTC)
agenta-documentation Ready Ready Preview, Comment Jun 23, 2026 4:03pm

Request Review

@dosubot dosubot Bot added size:XXL This PR changes 1000+ lines, ignoring generated files. feature frontend labels Jun 23, 2026
@coderabbitai

coderabbitai Bot commented Jun 23, 2026 •

Copy link
Copy Markdown
Contributor

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 2b0f93b9-27d9-4647-9533-11f6ffbf9f33

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

  • 🔍 Trigger review
✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create PR with unit tests
  • Commit unit tests in branch fe-feat/agent-playground-generation

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

Frontend layer for agent workflows, stacked on the agent chat slice (#4780).

Playground:
- AgentChatPanel as the third generation arm; buildAgentRequest + per-entity isAgentMode;
  multi-session tabs, attachments, conversation rewind; streaming over the v6 /messages
  contract (SSE).
- Rendering: markdown (tables, headings, code highlight, LaTeX), collapsible thinking,
  tool cards, streamed usage/cost, failed-run error surfaced from the trace.
- Session history picker; open-a-session-from-observability deep link.

Observability:
- Sessions table renders cell data (page-store context) + real height + id truncation.
- Agent reply shown in Last output, session-drawer conversation, and Traces Outputs
  (read the nested agent span, not the root's streamed-generator repr).
- Open-in-Playground replays an agent run from its root span (agent chat, not completion).

Agent config controls are taken verbatim from #4775; AgentConfigControl additionally
wires gatewayTools (Composio) into the tool picker (2-line delta on top of #4775).
@ardaerzin
ardaerzin force-pushed the fe-feat/agent-playground-generation branch from e61dd95 to 9955428 Compare June 23, 2026 16:02
@mmabrouk
mmabrouk changed the base branch from fe-feat/agent-chat-ui-slice to big-agents June 23, 2026 18:25
@mmabrouk
mmabrouk merged commit ff12a97 into big-agents Jun 24, 2026
22 of 29 checks passed

This branch was successfully deployed

1 active deployment
Preview — 99554286 Deployed Jun 23, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

feature frontend size:XXL This PR changes 1000+ lines, ignoring generated files.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants