feat(agent-service): enable Texera Agent to do workflow editing and execution#4540
Merged
Conversation
Introduce a standalone Bun/TypeScript service (agent-service) that hosts conversational agents backed by the Vercel AI SDK, plus the Angular panel and services that drive it over HTTP and a per-agent WebSocket. This slice contains the framework only: step tree with HEAD and checkout, per-step context assembly from the live DAG (via the compile API), auto-persist of workflow edits to the dashboard, and shared operator metadata loaded at startup. The agent has no tools yet, so it can converse about the workflow without editing or executing it. Workflow editing/execution tools, the SyncExecutionResource endpoint, and the single-node and k8s deployment configs ship in follow-up PRs.
Add Bun-native unit tests for the two pure-logic state modules (WorkflowState, WorkflowResultState) covering operator and link CRUD, sub-DAG walking, ancestor-path result lookup, and checkout semantics. Add an agent-service job to the build workflow that runs on Ubuntu, macOS, and Windows: install with bun, prettier check, tsc --noEmit, and bun test. Run prettier --write across existing source so the new format:check step is green.
Remove types/execution.ts, agent/workflow-result-state.ts (and its test), the computingUnitId field, and the result-summary plumbing in server.ts. The /operator-results route stays and returns an empty map so the frontend keeps working; branch-2 (tools) populates it once executeOperator lands. Refactor server.ts to expose a buildApp() factory and start()/import.meta gate so tests can drive the routes via app.handle(new Request(...)) — the standard Elysia testing pattern. Move .onError onto the agentsRouter instance so thrown errors inside its handlers are caught (Elysia 1.x scopes error handlers to the instance they are attached to). Replace DELETE's throw with a direct set.status = 404 for predictability. Add server.test.ts covering /health, agent create/list/get/delete, the stop/clear/operator-results control routes, settings PATCH round-trip, invalid JWT rejection, and missing-modelType validation.
Add **/.env.example, **/.prettierrc, and **/bun.lock to paths-ignore. These are config and lockfile artifacts that do not carry license headers, and skywalking-eyes flagged them in the agent-service tree.
Drop the step-id badge that rendered at the top-left of each chat message and the highlight-affected-operators toggle that rendered at the top-right. Together they backed a canvas overlay that drew a dashed region around all operators touched by a user message and a per-operator step badge stack. Removed plumbing: - agent.service: highlightedMessageIdSubject/$, setHighlightedMessage, getHighlightedMessageId, operatorStepsMapSubject/$, updateOperatorStepsMap, addStepRef, getOperatorStepRefs, getOperatorStepsMap, getOperatorsForMessage, extractOperatorIdsFromToolResult, extractOperatorIdsFromObject, tryParseJson - agent-types: OperatorStepRef interface - agent-chat: highlightedMessageId field, highlightedMessageId$ subscription, isMessageHighlighted, toggleMessageHighlight, the badge and toggle markup, .message-highlight-btn styles - workflow-editor: showStepBadges, stepBadges, messageRegionElement, highlightedMessageOperators, highlightedMessageId, currentOperatorStepsMap, handleStepBadges, updateStepBadgePositions, onStepBadgeClick, handleMessageRegion, updateMessageRegion, updateMessageRegionPath, the .step-badges-container/.step-badge/.message-region styles, and the step-badges template block
The service backed canvas diff highlighting for agent-applied actions (showDiff/clearDiff/toggleDiff via the version-diff helpers in WorkflowVersionService). Nothing renders agent actions on the canvas anymore, so drop the service and remove its single injection in WorkflowEditorComponent. The version-diff helpers it called remain; they are still used by WorkflowVersionService for the workflow version history feature.
Eight files in this branch had drift unrelated to the agent-service framework. Restore each to main: - dashboard/component/admin/user: drop sortByAffiliation/sortByJoiningReason columns - dashboard/service/user/workflow-version: drop highlightOpVersionDiffSimple and highlightOpBoundaryDashed (only callers were the now-removed AgentActionService) - dashboard/type/dashboard-entry: revert coverImageUrl signature tweak - workspace/component/menu: restore Status toggle, Edit Description button, region toggle, and onClickOpenShareAccess revoke-redirect logic - workspace/component/property-editor/operator-property-edit-frame: drop the editingTitle wrapper and the orphan .agent-interaction-controls styles (no markup consumed them)
- Delete bin/agent-service.dockerfile. The framework PR no longer ships a deploy artifact; the dockerfile belongs with the deploy PRs (single-node, k8s). - Restore bin/computing-unit-master.dockerfile (drop the LICENSE/NOTICE copy and application.ini removal — unrelated to agent-service). - Restore bin/litellm-config.yaml (drop the extra commented-out model entries; the proxy already accepts the upstream model list).
…t-fixtures.ts to match main Both diffs were unrelated to the agent-service framework — drift left behind on this branch from earlier work.
…e stubs The Edit / Save / Cancel / Reset-to-default buttons in the System Prompt modal called methods whose only effect was a "managed server-side" notification — the prompt is built from operator metadata at agent init and the local edit was never persisted. Same for the per-tool nz-switch, the Enable-all and Disable-all buttons. Drop the buttons + textarea from the template, the corresponding stub methods (saveSystemPrompt, resetSystemPromptToDefault, startEditingSystemPrompt, cancelEditingSystemPrompt, toggleToolEnabled, enableAllTools, disableAllTools, getEnabledToolsCount), the supporting fields (isEditingSystemPrompt, editingSystemPrompt), the unused `enabled: boolean` shape on availableTools, and the now-orphan SCSS rules (.settings-prompt-actions, .settings-textarea, .settings-info, .settings-option, .settings-tools-header and its nested .settings-tools-count / .settings-tools-actions, .settings-tool-item nz-switch). The System Prompt tab now shows the prompt read-only; the Tools tab shows the tool list read-only with name and description.
# Conflicts: # frontend/src/app/app.module.ts # frontend/src/app/workspace/component/agent-panel/agent-chat/agent-chat.component.ts # frontend/src/app/workspace/component/agent-panel/agent-panel.component.ts # frontend/src/app/workspace/component/agent/agent-panel/agent-chat/agent-chat.component.spec.ts # frontend/src/app/workspace/component/agent/agent-panel/agent-panel.component.html # frontend/src/app/workspace/component/workspace.component.ts
…tion endpoint Wire four tools onto TexeraAgent so the LLM can edit and run workflows: addOperator, modifyOperator, and deleteOperator (validated against the operator JSON schema registry), plus executeOperator, which compiles the current sub-DAG and runs it via the Texera execution service. Introduce /api/execution/:wid/:cuid/run on the JVM backend (SyncExecutionResource) as a synchronous endpoint that returns sampled output rows plus per-port shape metadata. Forward ENV_AUTH_JWT_SECRET to spawned computing-unit pods so they validate the agent's JWT. After each addOperator or modifyOperator, the service re-executes the affected operator and caches the result on the current step, so the model sees updated schemas and previews on its next step. Restores workflow-result-state, types/execution, and the computingUnitId plumbing that branch-1 dropped pending this branch (per the deletion commit's note: "branch-2 (tools) populates it once executeOperator lands"). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ng-zorro 21 dropped the nz-tabset selector. The agent-panel template was updated during the Angular 21 merge but this modal was missed, so clicking the info button next to the model name rendered an empty modal. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
github-actions
Bot
added
engine
dependencies
aglinxinyuan
approved these changes
Apr 28, 2026
aglinxinyuan
left a comment
aglinxinyuan
left a comment
Contributor
There was a problem hiding this comment.
Per discussion in PR #4495.
# Conflicts: # .github/workflows/github-action-build.yml # agent-service/src/agent/index.ts # agent-service/src/agent/texera-agent.ts # agent-service/src/agent/tools/index.ts # agent-service/src/api/execution-api.ts # agent-service/src/api/index.ts # agent-service/src/index.ts # agent-service/src/server.ts # agent-service/src/types/agent.ts # agent-service/src/types/index.ts # frontend/src/app/workspace/component/agent/agent-interaction/agent-interaction.component.html # frontend/src/app/workspace/component/agent/agent-interaction/agent-interaction.component.scss # frontend/src/app/workspace/component/agent/agent-interaction/agent-interaction.component.ts # frontend/src/app/workspace/component/agent/agent-panel/agent-chat/agent-chat.component.html # frontend/src/app/workspace/component/agent/agent-panel/agent-chat/agent-chat.component.scss # frontend/src/app/workspace/component/agent/agent-panel/agent-chat/agent-chat.component.ts # frontend/src/app/workspace/component/agent/agent-panel/react-step-detail-modal/react-step-detail-modal.component.html # frontend/src/app/workspace/component/agent/agent-panel/react-step-detail-modal/react-step-detail-modal.component.ts # frontend/src/app/workspace/component/workflow-editor/workflow-editor.component.html # frontend/src/app/workspace/component/workflow-editor/workflow-editor.component.ts # frontend/src/app/workspace/service/agent/agent.service.ts # frontend/src/app/workspace/service/copilot/copilot-prompts.ts
bobbai00
force-pushed
the
branch-2-prime-revised
branch
from
59f3abb to
235704f
Compare
github-actions
Bot
removed
dependencies
This was referenced Apr 28, 2026
yangzhang75
pushed a commit
to yangzhang75/texera
that referenced
this pull request
Jun 22, 2026
…xecution (apache#4540) <!-- Thanks for sending a pull request (PR)! Here are some tips for you: 1. If this is your first time, please read our contributor guidelines: [Contributing to Texera](https://github.com/apache/texera/blob/main/CONTRIBUTING.md) 2. Ensure you have added or run the appropriate tests for your PR 3. If the PR is work in progress, mark it a draft on GitHub. 4. Please write your PR title to summarize what this PR proposes, we are following Conventional Commits style for PR titles as well. 5. Be sure to keep the PR description updated to reflect all changes. --> ### What changes were proposed in this PR? <!-- Please clarify what changes you are proposing. The purpose of this section is to outline the changes. Here are some tips for you: 1. If you propose a new API, clarify the use case for a new API. 2. If you fix a bug, you can clarify why it is a bug. 3. If it is a refactoring, clarify what has been changed. 3. It would be helpful to include a before-and-after comparison using screenshots or GIFs. 4. Please consider writing useful notes for better and faster reviews. --> This PR adds the tools for the Texera Agent to do CRUD operations on workflow, and execute the workflow to see the execution result. ### Any related issues, documentation, discussions? <!-- Please use this section to link other resources if not mentioned already. 1. If this PR fixes an issue, please include `Fixes apache#1234`, `Resolves apache#1234` or `Closes apache#1234`. If it is only related, simply mention the issue number. 2. If there is design documentation, please add the link. 3. If there is a discussion in the mailing list, please add the link. --> Closes apache#4493 ### How was this PR tested? <!-- If tests were added, say they were added here. Or simply mention that if the PR is tested with existing test cases. Make sure to include/update test cases that check the changes thoroughly including negative and positive cases if possible. If it was tested in a way different from regular unit tests, please clarify how you tested step by step, ideally copy and paste-able, so that other reviewers can test and check, and descendants can verify in the future. If tests were not added, please describe why they were not added and/or why it was difficult to add. --> Manually tested ### Was this PR authored or co-authored using generative AI tooling? <!-- If generative AI tooling has been used in the process of authoring this PR, please include the phrase: 'Generated-by: ' followed by the name of the tool and its version. If no, write 'No'. Please refer to the [ASF Generative Tooling Guidance](https://www.apache.org/legal/generative-tooling.html) for details. --> No --------- Co-authored-by: Chen Li <chenli@gmail.com> Co-authored-by: Xinyuan Lin <xinyual3@uci.edu> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What changes were proposed in this PR?
This PR adds the tools for the Texera Agent to do CRUD operations on workflow, and execute the workflow to see the execution result.
Any related issues, documentation, discussions?
Closes #4493
How was this PR tested?
Manually tested
Was this PR authored or co-authored using generative AI tooling?
No