[integration] big-agents#4791
Open
mmabrouk wants to merge 435 commits into
Open
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Important Review skippedToo many files! This PR contains 1127 files, which is 977 over the limit of 150. To get a review, narrow the scope: Upgrade to a paid plan to raise the limit. ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: ⛔ Files ignored due to path filters (152)
📒 Files selected for processing (1127)
You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 10
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 76c33a7d-feff-4e5f-acc0-962498f74cfc
📒 Files selected for processing (70)
sdks/python/agenta/__init__.pysdks/python/agenta/sdk/agents/__init__.pysdks/python/agenta/sdk/agents/adapters/__init__.pysdks/python/agenta/sdk/agents/adapters/_runner_config.pysdks/python/agenta/sdk/agents/adapters/agenta_builtins.pysdks/python/agenta/sdk/agents/adapters/harnesses.pysdks/python/agenta/sdk/agents/adapters/in_process.pysdks/python/agenta/sdk/agents/adapters/local.pysdks/python/agenta/sdk/agents/adapters/sandbox_agent.pysdks/python/agenta/sdk/agents/adapters/vercel/__init__.pysdks/python/agenta/sdk/agents/adapters/vercel/messages.pysdks/python/agenta/sdk/agents/adapters/vercel/routing.pysdks/python/agenta/sdk/agents/adapters/vercel/sse.pysdks/python/agenta/sdk/agents/adapters/vercel/stream.pysdks/python/agenta/sdk/agents/dtos.pysdks/python/agenta/sdk/agents/errors.pysdks/python/agenta/sdk/agents/interfaces.pysdks/python/agenta/sdk/agents/mcp/__init__.pysdks/python/agenta/sdk/agents/mcp/errors.pysdks/python/agenta/sdk/agents/mcp/interfaces.pysdks/python/agenta/sdk/agents/mcp/models.pysdks/python/agenta/sdk/agents/mcp/parsing.pysdks/python/agenta/sdk/agents/mcp/resolver.pysdks/python/agenta/sdk/agents/mcp/wire.pysdks/python/agenta/sdk/agents/streaming.pysdks/python/agenta/sdk/agents/tools/__init__.pysdks/python/agenta/sdk/agents/tools/compat.pysdks/python/agenta/sdk/agents/tools/errors.pysdks/python/agenta/sdk/agents/tools/interfaces.pysdks/python/agenta/sdk/agents/tools/models.pysdks/python/agenta/sdk/agents/tools/parsing.pysdks/python/agenta/sdk/agents/tools/resolver.pysdks/python/agenta/sdk/agents/tools/wire.pysdks/python/agenta/sdk/agents/ui_messages.pysdks/python/agenta/sdk/agents/utils/__init__.pysdks/python/agenta/sdk/agents/utils/ts_runner.pysdks/python/agenta/sdk/agents/utils/wire.pysdks/python/agenta/sdk/decorators/routing.pysdks/python/agenta/sdk/engines/running/interfaces.pysdks/python/agenta/sdk/engines/running/utils.pysdks/python/agenta/sdk/middlewares/running/normalizer.pysdks/python/agenta/sdk/models/workflows.pysdks/python/agenta/sdk/utils/types.pysdks/python/agenta/tests/agents/test_streaming.pysdks/python/oss/tests/pytest/integration/agents/__init__.pysdks/python/oss/tests/pytest/integration/agents/test_transport_roundtrip.pysdks/python/oss/tests/pytest/unit/agents/__init__.pysdks/python/oss/tests/pytest/unit/agents/conftest.pysdks/python/oss/tests/pytest/unit/agents/golden/run_request.claude.jsonsdks/python/oss/tests/pytest/unit/agents/golden/run_request.pi.jsonsdks/python/oss/tests/pytest/unit/agents/golden/run_result.error.jsonsdks/python/oss/tests/pytest/unit/agents/golden/run_result.ok.jsonsdks/python/oss/tests/pytest/unit/agents/mcp/__init__.pysdks/python/oss/tests/pytest/unit/agents/mcp/test_resolver.pysdks/python/oss/tests/pytest/unit/agents/test_dtos_agent_config.pysdks/python/oss/tests/pytest/unit/agents/test_dtos_capabilities_events.pysdks/python/oss/tests/pytest/unit/agents/test_dtos_content_blocks.pysdks/python/oss/tests/pytest/unit/agents/test_dtos_harness_configs.pysdks/python/oss/tests/pytest/unit/agents/test_environment_lifecycle.pysdks/python/oss/tests/pytest/unit/agents/test_harness_adapters.pysdks/python/oss/tests/pytest/unit/agents/test_runner_adapter_config.pysdks/python/oss/tests/pytest/unit/agents/test_ui_messages.pysdks/python/oss/tests/pytest/unit/agents/test_wire_contract.pysdks/python/oss/tests/pytest/unit/agents/tools/__init__.pysdks/python/oss/tests/pytest/unit/agents/tools/test_models.pysdks/python/oss/tests/pytest/unit/agents/tools/test_parsing.pysdks/python/oss/tests/pytest/unit/agents/tools/test_resolver.pysdks/python/oss/tests/pytest/unit/test_normalizer_passthrough.pysdks/python/oss/tests/pytest/utils/test_messages_endpoint.pysdks/python/oss/tests/pytest/utils/test_routing.py
Comment on lines
+9
to
+13
| NOTE on packaging: the Node runner is NOT part of this Python wheel (``pip install agenta`` | ||
| stays pure Python; the wheel contains zero ``.ts``/``.js``). How a standalone Pi user obtains | ||
| the runner -- an ``npx`` npm package, a local checkout, or a Docker sidecar over HTTP -- is an | ||
| open distribution decision; see ``docs/design/agent-workflows/typescript-structure/``. Do NOT | ||
| silently bundle a JS runner into the wheel. |
There was a problem hiding this comment.
Align LocalBackend wording with the stated packaging contract.
Line 9-13 says the wheel must not bundle a JS runner, but Line 30 and the NotImplementedError messages still say “bundled JS”. This contradiction will confuse integrators.
Suggested wording fix
-class LocalBackend(Backend):
- """Run Pi (bundled JS) or Claude (``claude-agent-sdk``) on this machine."""
+class LocalBackend(Backend):
+ """Run Pi (external Node runner) or Claude (``claude-agent-sdk``) on this machine."""
...
raise NotImplementedError(
- "LocalBackend is not implemented yet (Phase 3: Pi via bundled JS, "
+ "LocalBackend is not implemented yet (Phase 3: Pi via external Node runner, "
"Phase 4: Claude via claude-agent-sdk)."
)
...
raise NotImplementedError(
- "LocalBackend is not implemented yet (Phase 3: Pi via bundled JS, "
+ "LocalBackend is not implemented yet (Phase 3: Pi via external Node runner, "
"Phase 4: Claude via claude-agent-sdk)."
)Also applies to: 30-38, 50-53
Comment on lines
+126
to
+136
| def __init__( | ||
| self, | ||
| *, | ||
| sandbox: str = "local", | ||
| url: Optional[str] = None, | ||
| command: Optional[Sequence[str]] = None, | ||
| cwd: Optional[str] = None, | ||
| timeout: float = float(os.getenv("AGENTA_AGENT_RUNNER_TIMEOUT_SECONDS", "180")), | ||
| ) -> None: | ||
| self._sandbox = sandbox | ||
| self._url = url |
There was a problem hiding this comment.
Validate sandbox at construction time.
Line 129 currently accepts any string; invalid values get sent over the wire and fail late. Restrict this to supported values (local, daytona) and raise a configuration error early.
Suggested validation
from ..dtos import (
@@
)
+from ..errors import AgentRunnerConfigurationError
@@
def __init__(
self,
*,
sandbox: str = "local",
@@
timeout: float = float(os.getenv("AGENTA_AGENT_RUNNER_TIMEOUT_SECONDS", "180")),
) -> None:
+ allowed_sandboxes = {"local", "daytona"}
+ if sandbox not in allowed_sandboxes:
+ raise AgentRunnerConfigurationError(
+ f"Unsupported sandbox '{sandbox}'. Expected one of: {sorted(allowed_sandboxes)}."
+ )
self._sandbox = sandbox
self._url = url
Comment on lines
+7
to
+20
| from agenta.sdk.agents.tools.models import MissingSecretPolicy | ||
|
|
||
| from .errors import MissingMCPSecretError | ||
| from .interfaces import MCPSecretProvider | ||
| from .models import MCPServerConfig, ResolvedMCPServer | ||
|
|
||
|
|
||
| class MCPResolver: | ||
| def __init__( | ||
| self, | ||
| *, | ||
| secret_provider: MCPSecretProvider, | ||
| missing_secret_policy: MissingSecretPolicy = MissingSecretPolicy.ERROR, | ||
| ) -> None: |
There was a problem hiding this comment.
Breaks declared layer direction by importing tools model into MCP.
MCPResolver currently depends on agenta.sdk.agents.tools.models.MissingSecretPolicy, but this cohort declares tools as depending on MCP, not the other way around. This reverse edge can create import-order fragility and circular dependency risk as the stack evolves. Move MissingSecretPolicy to a neutral/shared module (or MCP/shared contract module) and import it from both subsystems.
Possible direction
- from agenta.sdk.agents.tools.models import MissingSecretPolicy
+ from agenta.sdk.agents.shared.missing_secret_policy import MissingSecretPolicy(then define/move the enum in that shared module and update tools imports accordingly)
Comment on lines
+67
to
+75
| out = stdout.decode("utf-8", "replace") | ||
| err = stderr.decode("utf-8", "replace") | ||
| if not out.strip(): | ||
| raise RuntimeError( | ||
| f"Agent runner returned no output. exit={proc.returncode} stderr={err[-2000:]}" | ||
| ) | ||
| try: | ||
| return json.loads(out) | ||
| except json.JSONDecodeError as exc: |
There was a problem hiding this comment.
Treat non-zero subprocess exit as transport failure even with parseable JSON.
Line 74 returns parsed JSON without checking proc.returncode; a crashed runner can look successful if it emitted partial/legacy JSON before exiting non-zero.
Suggested fix
@@ async def deliver_subprocess(...):
out = stdout.decode("utf-8", "replace")
err = stderr.decode("utf-8", "replace")
+ if proc.returncode not in (0, None):
+ raise RuntimeError(
+ "Agent runner exited non-zero. "
+ f"exit={proc.returncode} stderr={err[-2000:]} stdout={out[:500]}"
+ )
if not out.strip():
raise RuntimeError(
f"Agent runner returned no output. exit={proc.returncode} stderr={err[-2000:]}"
)📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| out = stdout.decode("utf-8", "replace") | |
| err = stderr.decode("utf-8", "replace") | |
| if not out.strip(): | |
| raise RuntimeError( | |
| f"Agent runner returned no output. exit={proc.returncode} stderr={err[-2000:]}" | |
| ) | |
| try: | |
| return json.loads(out) | |
| except json.JSONDecodeError as exc: | |
| out = stdout.decode("utf-8", "replace") | |
| err = stderr.decode("utf-8", "replace") | |
| if proc.returncode not in (0, None): | |
| raise RuntimeError( | |
| "Agent runner exited non-zero. " | |
| f"exit={proc.returncode} stderr={err[-2000:]} stdout={out[:500]}" | |
| ) | |
| if not out.strip(): | |
| raise RuntimeError( | |
| f"Agent runner returned no output. exit={proc.returncode} stderr={err[-2000:]}" | |
| ) | |
| try: | |
| return json.loads(out) | |
| except json.JSONDecodeError as exc: |
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
dosubot
Bot
added
size:XXL
…kdown editor Three improvements to the toolbar'd markdown editor used by the Instructions and Skill drawers. - Link button now opens a popover asking for the URL (seeded with the existing link when the caret is on one) with Add/Update + Remove, instead of blindly applying the literal "https://". - Table controls: a size-picker popover inserts a table (rows×cols hover grid); when the caret is inside a table the same control becomes a menu for insert/ delete row and column + delete table. The Lexical engine already registered TablePlugin + the markdown table transformers — this just exposes them. - The editor now accepts a dropped Markdown file (.md/.markdown/.mdx/.txt or any text/* file): it reads the file text and replaces the content, with a drop overlay. Intercepts only file drags in the capture phase, so Lexical's own internal text drag/drop is untouched.
Apply the "comfortable" prose direction to how Markdown documents (AGENTS.md / SKILL.md) render, in both the live editor and the read-only preview. The shared Lexical editor theme rendered document headings far too large (h1 24px / 600 weight), gave fenced code blocks no block styling, and left tables/blockquotes inconsistent. Rather than change that theme globally (the prompt and chat editors share it), scope the new prose styles to the document editors via a `md-prose` wrapper class: - editor-theme.css: a `.md-prose` block with a calmer heading scale (17/14/13, 500 weight), bordered/rounded code blocks, an accent-rule italic blockquote, and roomier line-height. Uses antd semantic tokens so it adapts to dark mode. - MarkdownEditor: tag its rendered containers with `md-prose` (covers both the edit and preview panes, which both render through this editor). - MarkdownPreview: bring its marked/DOMPurify MD_CLASS in line with the same scale, plus table styling it previously lacked.
Follow-up to live testing of the document editor. - Headings: replace the single H2 button with a block-type menu (Normal text, Heading 1-3, Quote, Code block) that reflects the caret's current block. The editor only ever offered one heading level before. "Code block" inserts a real fenced block (the standard @lexical/code CodeNode, already registered in rich mode) instead of the inline-code button, which had made multi-line "code" read as a stack of inline chips. - Headings now render with a clearer hierarchy (20/16/14, 600 weight) so they stand apart from body text. - Lists: the shared theme gave ol and ul different left margins and tall items, so the two indented inconsistently. Normalize both to one padding + tight rows under .md-prose. - Table size-picker: its grid cells used a fixed light hex that vanished in dark mode. Switch to antd semantic tokens so the cells are visible in both themes. - Code block wraps (pre-wrap) and keeps its bordered/rounded surface.
…ghting The "Code block" option turned a multi-line selection into one code node per line via $setBlocksType — which rendered as a stack of inline-code chips with no block background, not a single block. Mirror the Lexical playground: on a range selection, insert one code node and write the text back with insertRawText so line breaks become code lines (collapsed selection still uses $setBlocksType). Also wire up code highlighting, which the rich editor never enabled: - MarkdownEditor registers `registerCodeHighlighting` (the CodeNode/ CodeHighlightNode types were registered but the highlighter was never turned on, so blocks rendered as plain monospace). Token colors come from the existing `editor-token*` theme classes (light + dark). - The toolbar shows a searchable language picker (`getCodeLanguageOptions`) when the caret is inside a code block, writing `CodeNode.setLanguage` so Prism highlights the right grammar.
`.editor-inner:not(.code-editor) .editor-code` (0-3-0) was overriding the document code-block padding/fill from `.md-prose .editor-code` (0-2-0), so blocks rendered with 8px padding instead of the intended 12px/14px. Scope the rule to the rendered input (`.editor-input:not(.markdown-view)`) to reach 0-4-0 and win, without touching the markdown-source view. Verified live: a created code block now computes padding 12px 14px with the bordered/rounded surface.
Two changes from live testing of the document editor. - Drop the inline-code (`</>`) toolbar button. It applied per-line inline-code formatting, which read as a stack of chips and was the thing being mistaken for a code block. Code now means one thing: the block-type menu's "Code block". - Move the language picker out of the toolbar onto the code block itself. A new CodeBlockLanguageMenu renders a floating language Select pinned to the top-right of the code block the caret is in (fixed-position portal, recomputed on editor update + scroll/resize), driving CodeNode.setLanguage for Prism highlighting. The block gets 40px top padding to seat it. Verified live in the running app: toolbar no longer shows an inline-code button; selecting lines + "Code block" yields one styled block with the language picker seated on its top-right, aligned to the block and tracking scroll.
…polish it
The code block carried two language indicators that overlapped: the editor's
built-in uppercase `::after` label (globals.css, `attr(data-highlight-language)`)
and the new interactive picker. Hide the built-in label on document code blocks
(0-6-1 selector to beat globals' 0-5-1) so only the picker shows.
Polish the picker: borderless + compact (24px control, 11px text, sans), and
render the friendly language name ("JavaScript") for the selected value instead
of the raw id.
Verified live in the running app: two code blocks in one editor render
independently; the picker seats on the active block's top-right with no overlap;
the built-in `::after` label computes to `none`.
…e focused one Each code block reserves a 40px top strip for the language picker, but the picker only rendered on the block the caret was in — so other blocks showed an empty gap. Render a picker for every code block in the document (iterating the root's CodeNodes rather than the selection), clipped to the editor's scroll viewport so blocks scrolled out of view don't float a picker over the toolbar. Read-only panes get the picker too (disabled), so previews also fill the strip. Verified live: with two code blocks and the caret in a heading (neither focused), both blocks render their picker aligned to their own top-right.
…ions
Triggers showed a plain sentence ("No triggers yet. Add an app trigger or a
schedule…") while Tools/MCP/Skills use "No X yet — <add link>". Match it:
"No triggers yet — add a trigger", where the link opens the existing add menu
(App trigger / Scheduled trigger), so the "app trigger or a schedule" intent is
kept in the menu rather than the copy.
To guarantee parity, extract the shared `AddTextLink` (was local to
AgentTemplateControl) into its own module and use it in both. AddTriggerDropdown
gains an optional `trigger` so the empty-state link reuses the header's menu.
Verified live: the Triggers link is byte-identical in class to the Skills link
and opens the same add menu.
… playground
Nested numbered lists rendered every level as decimal (1., 1.), so depth was
invisible. Lexical already DOM-nests indented items (the wrapper li carries
`editor-nested-list-item`, which is marker-less), and the markdown round-trip
preserves nesting at a 4-space indent (LIST_INDENT_SIZE). So this is purely a
marker-styling gap: add depth-cycling `list-style-type` via descendant selectors
like the playground — ordered 1 → A → a → I → i, unordered • → ◦ → ▪ — scoped to
the rendered document view, plus the matching rules in MarkdownPreview.
Verified live: a nested ordered item computes upper-alpha ("A.") under the
decimal parent, with the wrapper li marker-less.
… font - On a cold open the drawer is still animating and the markdown still hydrating, so the first rect read is stale and the picker only appeared once a later event (e.g. a hover) nudged a recompute. Re-run recompute across the next frame + a few timeouts so it settles on the real position without interaction. - The picker portals to <body>, outside the antd theme container, so antd's inner `.ant-select-selection-item` fell back to a serif (Times). Set the app font (`var(--ant-font-family)`) on the wrapper and force the inner item / placeholder / search input to inherit it. Verified live: now renders Inter.
…rkers CR #4915: - The file-drop handler was memoized on `[dropEnabled]` with an eslint-disable, so it could call a stale `onChange`. That's reachable: SkillFormView passes a fresh inline `onChange={(v) => set("body", v)}` every render. Make `handleChange` a `useCallback([onChange])`, hoist the pure drop predicates to module scope, and add `handleChange` to the drop deps — the disable is gone and exhaustive-deps is clean. - MarkdownPreview only carried ordered nesting to lower-alpha while the editor CSS continues to upper-roman / lower-roman, so depth 4-5 lists rendered differently in the preview. Add the missing two levels.
feat(frontend): markdown document editor — toolbar, code blocks, nested lists, rendering
…s (phase 1) Pure, verbatim move out of the 2070-line AgentTemplateControl into a new `agentTemplate/` folder — no logic change: - agentTemplateUtils.ts — enumLabel, countSummary, cloneItem - itemDescriptors.tsx — ItemDescriptor + describeTool/Mcp/Skill/Instruction and their classifiers (toolName, isFunctionTool, isStaticSkill, isEmbedRefSkill, mdPreview, …) - ItemRow.tsx — ItemAvatar, ItemRow, InstructionsFileRow Orchestrator re-imports them; dropped the now-dead phosphor/toolUtils imports. 2070 → 1628 lines. tsc + lint clean.
…ry (phase 2) Tools, MCP servers and skills were three near-identical copies of: a list body, a draft-then-save drawer, and the editing/commit/validate state. Collapse them onto one code path, driven by a per-kind registry: - itemKinds.tsx — ITEM_KINDS: field, describe, FormView, drawerTitle/width, editView, jsonOnly, isReadOnly, createSeed, draftInvalid per kind. - useConfigItemDrawer.ts — the shared editing/draft/commit/remove/validate machine, writing to each kind's array via the registry. - ConfigItemList.tsx — one list body (rows + empty state) for all three. - The three ConfigItemDrawer blocks become one registry-driven render. Behavior preserved exactly (each per-kind rule mapped 1:1). Orchestrator 1638 → 1423 lines. tsc + lint clean.
…Harness Move the agent-template model/connection/harness/sandbox state and the Model & harness + Advanced section JSX out of AgentTemplateControl into a dedicated useModelHarness hook. The two sections share one coupled state machine (the model/connection state feeds both), so they live together and the hook returns the summaries + section bodies the orchestrator renders. AgentTemplateControl drops from ~1.4k to 672 lines; behavior is unchanged.
Bump lexical core + all @lexical/* packages from ^0.40.0 to ^0.46.0
across oss, ee, and the @agenta/{ui,entities,entity-ui} packages
(plus the root @lexical/eslint-plugin).
The only code change required across the editor (10 custom nodes, 12
extensions, plugins, markdown transformers) is one mechanical fix for
the 0.46 removal of unsafe type parameters on node-traversal methods:
TableCellResizerPlugin used tableRow.getChildren<TableCellNode>(), now
getChildren() as TableCellNode[].
Move the agent-template Tools add/remove logic (inline function tools, builtin/gateway tools, async workflow-reference tools, and the derived selected-name set + referenceable-workflow pool) out of AgentTemplateControl into a useAgentTools hook. The orchestrator now just wires the handlers into the picker and renders. AgentTemplateControl drops to 580 lines (from ~2.1k before the split); behavior is unchanged.
Editing across a fenced code block with mixed { / {{ braces while
{{tokens}} exist could crash the editor with React "Maximum update
depth exceeded". Pre-existing (reproduces on 0.40), surfaced during
lexical upgrade testing.
Root cause: TokenTypeaheadPlugin called React setState (setAnchor/
setInputQuery) synchronously inside a Lexical registerUpdateListener. A
burst of commits drove setState → re-render → commit past React's depth
limit. A contributing factor was TokenPlugin.$transformNode scheduling
a nested editor.update() from inside a node transform to reposition the
caret, spawning extra commits.
Fix:
- TokenTypeaheadPlugin: compute the anchor synchronously but apply React
state in a coalesced queueMicrotask, identity-stable, never inside the
commit — no commit burst can exceed React's update-depth limit.
- TokenPlugin: reposition the caret synchronously via navigateCursor
instead of a nested editor.update() (behaviorally identical).
Verified: crash recipe + heavy stress no longer crash; auto-close and
manual-close caret behavior unchanged.
…nnel Addresses a review finding: after selecting a suggestion, the deferred microtask could re-anchor and reopen the menu right after selectOption closed it (the listener re-anchors on the caret position selectOption sets, and the deferred flush ran after the synchronous setAnchor(null)). Unify all anchor/query writers (update listener, selection, escape, click-outside) behind a single scheduleTypeahead → flushTypeahead channel. The last write before the microtask wins, so selectOption's close — which runs after the listener — wins by construction. Keeps the React update out of the synchronous Lexical commit (crash guard) and removes the need for ad-hoc suppression flags.
The lifted flushTypeahead used a component-scoped mountedRef (init true, set false on unmount) that was never reset in effect setup. Under React StrictMode's mount → cleanup → remount cycle it stayed false, so the flush bailed permanently in dev and the typeahead menu could stop opening. On React 19 a useState setter after unmount is a harmless no-op, so the guard was unnecessary machinery — removed it rather than patching the reset, eliminating the StrictMode footgun.
@agenta/ui supports React >=18.0.0, so the rationale shouldn't pin to React 19. setState-after-unmount is a no-op on the supported range; keep the StrictMode/microtask reasoning without the version-specific claim.
Address review nitpicks: move the item-avatar's static font styles into Tailwind classes (only background stays inline/dynamic), and compress the extracted modules' multi-line narrative comments to one-/two-line notes, keeping the genuinely surprising constraints (deliberate model non-clearing, vault-async slug, tracked harness-capabilities gap). No behavior change.
chore(frontend): upgrade Lexical 0.40 → 0.46
junaway
changed the title
Clarify in itemKinds that tool creation seeds from the picker, not the registry's createSeed stub (defuses a false-positive review read of an empty-tool save path). Record two review items from PR #4923 as deferred against the upcoming schema-driven config redesign: tighter per-kind tool draft validation (kind discriminated union) and re-binding the named-connection slug on a provider change (gated on first-class provider via ModelSpec). Both are correct for the redesign, not the current code, so they live in the design doc's Deferred list.
…ate-control refactor(frontend): split AgentTemplateControl into focused modules
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
5 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.
Context
big-agentsis the integration branch for the agent-workflows feature. Every agent PR targetsbig-agents(directly, or by stacking on one that does). The plan is to review and merge each sub-PR intobig-agents, then mergebig-agentsintomainas a single unit.This PR is a draft tracker. It stays open until all the open sub-PRs below are merged into
big-agents. The branch started from an empty commit, so the diff fills in as sub-PRs land.Integrated PRs
Each box gets checked when that PR is merged into
big-agents. Indented items stack on the item above them.SDK and service
Runner
big-agents(the relay-bug fix, the CI job, and a superset of its tests already landed via feat(agent): runner engines, HTTP server, tracing, and docker image #4778 + chore(agent): make sandbox-agent runner first-class #4786)Frontend
Hosting
Sandbox-agent deployment
Docs
Branch-only (no PR yet)
These design-doc branches are stacked on
big-agentsbut have no PR. Open one if you want them reviewed separately, otherwise they fold in with the docs.docs/agent-model-config-and-provider-authdocs/agent-skills-configdocs/agent-code-tool-sandboxdocs/agent-harness-capabilitiesNotes
big-agents(feat(agent): runner engines, HTTP server, tracing, and docker image #4778 + chore(agent): make sandbox-agent runner first-class #4786 already carry its tests, CI job, and relay-bug fix; itsversion.tswas stale["pi","rivet"]).big-agentsas chore(railway): add sandbox-agent preview deployment #4802 / chore(kubernetes): deploy sandbox-agent sidecar #4803 / ci(agent): build and test sandbox-agent images #4804.