fix(eve): render background task turns in the dev TUI - #1561
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Bundle + Package Summary:
|
| Area | Metric | Baseline | Current | Delta |
|---|---|---|---|---|
| Package | Packed tarball | 7.68 MB | 7.67 MB | -2.4 kB ✅ |
| Package | Unpacked publish size | 28.76 MB | 28.77 MB | +9.9 kB |
| Package | Installed footprint | 72.09 MB | 72.10 MB | +9.9 kB |
| Package | Published files | 3068 | 3072 | +4 |
| Package | Installed files | 6767 | 6771 | +4 |
| Runtime | Unique function payloads | 2 | 2 | 0 |
| Runtime | Total function bytes | 17.13 MB | 17.14 MB | +12.3 kB |
| Runtime | Public routes | 13 | 13 | 0 |
Changed function payloads vs tasks/05-send (62c2f65) (2)
| Function | Status | Baseline | Current | Delta | Route changes |
|---|---|---|---|---|---|
functions/__server.func |
changed | 8.57 MB | 8.57 MB | +6.1 kB |
none |
functions/.well-known/workflow/v1/flow.func |
changed | 8.57 MB | 8.57 MB | +6.1 kB |
none |
eve init install
| Metric | Baseline | Current | Delta |
|---|---|---|---|
| Installed footprint | 110.50 MB | 110.51 MB | +9.9 kB |
| Installed packages | 121 | 121 | 0 |
| dependencies | 4 | 4 | 0 |
| devDependencies | 2 | 2 | 0 |
| Dependency package bytes | 43.17 MB | 43.18 MB | +9.9 kB |
| devDependency package bytes | 5.04 MB | 5.04 MB | 0 B ➖ |
Build Metadata
- Preset:
vercel - Nitro:
nitro@3.0.260610-beta - Output directory:
apps/fixtures/weather-agent/.vercel/output - Build metadata timestamp: 2026-08-05T19:02:27.688Z
- Route aliases: 13 public, 1 internal (14 total aliases)
- Vercel routes in config: 16
- Severity legend: 🔴 dominant/large, 🟠 notable, 🟡 watch, ⚪ small
Package Drill-Down
Package Details
- Package:
eve@0.30.7 - Package directory:
packages/eve - Tarball: 7.67 MB (
eve-0.30.7.tgz) - Unpacked payload: 28.77 MB across 3072 published files
- Installed footprint: 72.10 MB across 6771 installed files
- Installed root package: 27.42 MB
- Installed dependencies: 44.68 MB
- Runtime dependencies: 2
- Peer dependencies: 5 (4 optional)
Installed footprint is measured from an isolated temporary npm install of the packed tarball.
Heavy installed dependencies
eve: 27.42 MB (38.0%)@rolldown/binding-linux-x64-gnu: 19.35 MB (26.8%)ai: 6.60 MB (9.2%)zod: 5.07 MB (7.0%)undici: 3.50 MB (4.9%)
Publish payload breakdown
Published file size
🔴 dist/src/compiled/shadcn-registry/index.js [################........] 9.68 MB 33.7%
🟠 dist/src/compiled/@photon-ai/chat-adapter-ime... [####....................] 2.42 MB 8.4%
🟠 dist/src/compiled/experimental-ai-sdk-code-mo... [###.....................] 1.51 MB 5.2%
🟡 dist/src/compiled/_chunks/node/undici-DWL_MYm... [#.......................] 502.4 kB 1.7%
🟡 dist/src/compiled/_chunks/workflow/undici-DWL... [#.......................] 502.4 kB 1.7%
🔴 Other published files [########################] 14.16 MB 49.2%
Installed footprint breakdown
Installed package size
🔴 eve [########################] 27.42 MB 38.0%
🔴 @rolldown/binding-linux-x64-gnu [#################.......] 19.35 MB 26.8%
🔴 ai [######..................] 6.60 MB 9.2%
🔴 zod [####....................] 5.07 MB 7.0%
🟠 undici [###.....................] 3.50 MB 4.9%
🟠 nitro [##......................] 2.41 MB 3.3%
🔴 Other installed packages [#######.................] 7.76 MB 10.8%
Runtime dependencies (2)
| Package | Range | Notes |
|---|---|---|
nitro |
3.0.260610-beta |
|
undici |
8.9.0 |
Peer dependencies (5)
| Package | Range | Notes |
|---|---|---|
@opentelemetry/api |
^1.0.0 |
optional peer |
ai |
catalog: |
|
braintrust |
^3.0.0 |
optional peer |
just-bash |
^3.0.0 |
optional peer |
microsandbox |
^0.5.0 |
optional peer |
eve init install drill-down
eve init install details
- Command:
eve init my-agent - Package manager:
npm - Installed footprint: 110.51 MB across 8647 installed files
- Installed packages: 121 total (115 transitive-only)
- dependencies: 4 direct packages totaling 43.18 MB
- devDependencies: 2 direct packages totaling 5.04 MB
- Other transitive package files: 62.29 MB
Installed footprint is measured from an isolated temporary eve init my-agent using the current packed eve tarball.
Heavy installed dependencies
@typescript/typescript-linux-x64: 27.95 MB (25.3%)eve: 27.42 MB (24.8%)@rolldown/binding-linux-x64-gnu: 19.35 MB (17.5%)zod: 9.02 MB (8.2%)ai: 6.60 MB (6.0%)
Installed footprint breakdown
Installed package size
🔴 @typescript/typescript-linux-x64 [########################] 27.95 MB 25.3%
🔴 eve [########################] 27.42 MB 24.8%
🔴 @rolldown/binding-linux-x64-gnu [#################.......] 19.35 MB 17.5%
🔴 zod [########................] 9.02 MB 8.2%
🔴 ai [######..................] 6.60 MB 6.0%
🟠 undici [###.....................] 3.50 MB 3.2%
🔴 Other installed packages [##############..........] 16.67 MB 15.1%
dependencies (4)
| Package | Range | Installed size | Share |
|---|---|---|---|
@vercel/connect |
0.4.3 |
141.0 kB | 0.1% |
ai |
^7.0.38 |
6.60 MB | 6.0% |
eve |
file:eve-0.30.7.tgz |
27.42 MB | 24.8% |
zod |
4.4.3 |
9.02 MB | 8.2% |
devDependencies (2)
| Package | Range | Installed size | Share |
|---|---|---|---|
@types/node |
24.x |
2.54 MB | 2.3% |
typescript |
7.0.2 |
2.50 MB | 2.3% |
Function Drill-Down
Payload Size Graph
Unique function payload size and share of total
🔴 functions/.well-known/workflow/v1/flow.func [########################] 8.57 MB 50.0%
🔴 functions/__server.func [########################] 8.57 MB 50.0%
Top Function Payloads
🟠 functions/.well-known/workflow/v1/flow.func • 1 public route • 8.57 MB
| Metric | Value |
|---|---|
| Public routes | /.well-known/workflow/v1/flow |
| Runtime | nodejs24.x |
| Handler | index.mjs |
| Payload | 8.57 MB |
| Function files | 8.57 MB across 44 files |
| Traced dependencies | 0 B |
| Signal | 🟠 Bundled file index.mjs is 2.38 MB (27.8%) |
🟠 🔎 Dependency Analysis
📦 Bundled files:
Bundled file size
🟠 index.mjs [########################] 2.38 MB 27.8%
🟠 _chunks/runtime-artifacts.mjs [################........] 1.63 MB 19.0%
🟡 _libs/undici.mjs [##########..............] 980.5 kB 11.4%
🟡 _chunks/sandbox.mjs [########................] 769.0 kB 9.0%
🟡 _libs/@ai-sdk/gateway+[...].mjs [####....................] 432.8 kB 5.0%
🟠 Other bundled files [########################] 2.38 MB 27.8%
🧾 Vercel Config
{
"handler": "index.mjs",
"launcherType": "Nodejs",
"shouldAddHelpers": false,
"supportsResponseStreaming": true,
"runtime": "nodejs24.x",
"maxDuration": "max",
"experimentalTriggers": [
{
"type": "queue/v2beta",
"topic": "__eve776561746865722d6167656e74_wkf_workflow_*",
"consumer": "default",
"retryAfterSeconds": 5,
"initialDelaySeconds": 0
}
],
"environment": {
"WORKFLOW_PRECONDITION_GUARD": "1"
}
}🟠 functions/__server.func • 12 public routes, 1 internal alias • 8.57 MB
| Metric | Value |
|---|---|
| Public routes | //eve/v1/callback/[token]/eve/v1/connections/[name]/callback/[token]/eve/v1/health/eve/v1/info/eve/v1/session/eve/v1/session/[sessionId]/eve/v1/session/[sessionId]/cancel/eve/v1/session/[sessionId]/stream/eve/v1/session/clear/eve/v1/session/compact/eve/v1/session/reset |
| Internal aliases | /__server |
| Runtime | nodejs24.x |
| Handler | index.mjs |
| Payload | 8.57 MB |
| Function files | 8.57 MB across 44 files |
| Traced dependencies | 0 B |
| Signal | 🟠 Bundled file index.mjs is 2.38 MB (27.8%) |
🟠 🔎 Dependency Analysis
📦 Bundled files:
Bundled file size
🟠 index.mjs [########################] 2.38 MB 27.8%
🟠 _chunks/runtime-artifacts.mjs [################........] 1.63 MB 19.0%
🟡 _libs/undici.mjs [##########..............] 980.5 kB 11.4%
🟡 _chunks/sandbox.mjs [########................] 769.0 kB 9.0%
🟡 _libs/@ai-sdk/gateway+[...].mjs [####....................] 432.8 kB 5.0%
🟠 Other bundled files [########################] 2.38 MB 27.8%
🧾 Vercel Config
{
"handler": "index.mjs",
"launcherType": "Nodejs",
"shouldAddHelpers": false,
"supportsResponseStreaming": true,
"runtime": "nodejs24.x"
}Build Timing: e2e/fixtures/agent-tools-sandbox
This is an informational timing measurement inside eve build, from preflight through publication. Output-size measurement and profile writing are excluded.
Build mode: deployable Vercel build with sandbox template prewarm included.
- Build pipeline: 2.68 s -> 2.01 s (-670.9 ms) vs
tasks/05-send (62c2f65). - Timing is informational: shared GitHub runners are too variable for a hard timing budget.
Detailed phase timings vs `tasks/05-send (62c2f65)`
| Phase | Baseline | Current | Delta |
|---|---|---|---|
extension.check |
8.9 ms | 1.0 ms | -7.9 ms |
project.resolve |
5.9 ms | 0.4 ms | -5.5 ms |
workspace.create |
1.5 ms | 0.5 ms | -1.0 ms |
host.prepare |
267.2 ms | 198.7 ms | -68.5 ms |
vercel.service-prefix.resolve |
1.9 ms | 1.7 ms | -0.2 ms |
nitro.create |
174.6 ms | 200.8 ms | +26.2 ms |
sandbox.prewarm |
245.6 ms | 218.6 ms | -27.0 ms |
nitro.cache.prepare |
0.2 ms | 0.2 ms | 0.0 ms |
nitro.prepare |
0.8 ms | 0.8 ms | 0.0 ms |
nitro.public-assets |
0.5 ms | 0.5 ms | 0.0 ms |
nitro.prerender |
0.4 ms | 0.3 ms | -0.1 ms |
nitro.bundle |
1.79 s | 980.8 ms | -809.1 ms |
nitro.cache.write |
0.2 ms | 0.2 ms | 0.0 ms |
vercel.workflow-function.materialize |
177.0 ms | 209.6 ms | +32.6 ms |
agent-summary.emit |
0.4 ms | 0.5 ms | +0.1 ms |
nitro.close |
0.1 ms | 0.1 ms | 0.0 ms |
output.publish |
2.8 ms | 192.4 ms | +189.6 ms |
workspace.remove |
1.7 ms | 1.8 ms | +0.1 ms |
|
Tasks 06 through 09 are coalesced in this PR as one background-task TUI rendering changeset. |
Adds the packages/eve/src/tasks module cluster for experimental.tasks: the TaskStatus/TaskView/TaskOutput contract, the pure lifecycle transition function (terminal is final, failed carries its error output, working <-> input_required), the eve.tasks session index that holds each task's private command token, and replay-idempotent task ids derived from the agent-handle operation-id derivation. The durable task run is a small stable-named workflow that is the single writer for one task's lifecycle: it consumes commands over a private hook, applies the transition function, and appends a full TaskView snapshot per accepted command to its eve.task run stream. Node-side controls (start, command, tail-read, ready-wait) compose inside dispatch and tool steps. Nothing references these modules yet; no runtime behavior changes. Signed-off-by: Rui Conti <ruiconti@gmail.com>
Injects task_peek, task_await, task_cancel, and task_sleep into the harness tool map only when the root agent sets experimental.tasks, following the implicit agent-tool pattern: inline definitions, authored shadowing, disableTool support, session-shape hiding for self-delegated children, and dispatch-time re-validation to come with the delegated execution mode. task_peek/task_await/task_cancel are execute-less runtime actions with the new task-control kind — they need durable session state and world access, so the runtime-action dispatch step will execute them. task_sleep records the existing durable turn sleep in-loop. Dynamic workflow interrupts reject task-control tools explicitly. Nothing dispatches these actions yet; without the flag no tool set, prompt, or event stream changes. Signed-off-by: Rui Conti <ruiconti@gmail.com>
With the flag on, every subagent dispatch (fresh start, agentId continuation, local and remote) becomes delegated execution: - the durable task run is created before the child dispatch side effect, so a fast child always finds a live command hook; a replay re-derives the same deterministic command token and the duplicate run loses the hook claim; - the child's reply address is the task run's private hook instead of the parent turn's inbox, so the existing local notification, HITL forwarding, and remote callback wires deliver to the task unchanged; - the originating call resolves immediately with a task receipt whose parked outcome settles the agent handle through the existing resolve path, keeping history provider-valid with exactly one result; - a dispatch that never acknowledges a child fails its task out of band and returns the failure directly, so the model never sees a task id for work that never started. The task run translates the child wire into lifecycle commands (settled turn outcomes to complete/fail/cancel, HITL batches to input_required, authorization events to blocked/unblocked) and wakes the parent through the ordinary session delivery path when a task becomes ready — a parked parent starts a turn, an active turn observes the notification at its next safe boundary. task_peek and task_cancel execute inside the dispatch step, which owns the session task index and world access; task_cancel commits cancelled before propagating a cooperative abort through the handle address. task_await rides the turn's existing inbox wait: a small aggregation run polls the selected snapshot streams and posts the one tool-result the pending key expects, exiting quietly when the waiting turn is gone. Parent finalization cancels live tasks before terminating children. The flag also implies conversation-mode children so experimental.tasks and experimental.subagentPersistentSessions never produce a third mode. Child usage accounting for delegated tasks is deferred; the subagent-start machinery moved to its own module to keep the dispatch step within the file-length cap. Without the flag, dispatch behavior, wire shapes, and events are unchanged (agent-messaging scenario suite passes unmodified). Signed-off-by: Rui Conti <ruiconti@gmail.com>
Signed-off-by: Rui Conti <ruiconti@gmail.com>
task_send completes the parent-side task controls:
- an input_required task accepts an inputResponses batch, delivered to
the parked child session with the task run as the reply target, and
returns to working;
- a terminal task accepts a follow-up message, which runs the full
delegated continuation through the agent handle: a new task on the
same child session, returning the new receipt;
- a working task surfaces AGENT_BUSY instead of queueing (settled
decision; queueing on the task run is the reversible follow-up).
Sends verify ownership through the session task index, and the
continued handle is parked directly because task_send's tool-result
never flows through the handle-settling resolve path. The task-control
executors split into execution/tasks/{delegate,control-shared,send} to
keep each file within the length cap. Includes the release changeset
for the experimental.tasks feature.
Signed-off-by: Rui Conti <ruiconti@gmail.com>
Signed-off-by: Rui Conti <ruiconti@gmail.com>
Signed-off-by: Rui Conti <ruiconti@gmail.com>
9624f19 to
a63a12d
Compare
8374295 to
11528be
Compare
11528be to
a63a12d
Compare
Stacked on #1522. Background tasks exposed four related failures in the dev TUI:
task_awaitand the completion wake could report the same ready transition in two parent turns.These are one rendering lifecycle: a background child outlives its originating parent turn, wakes the parent independently, and finishes at its own stream boundary.
Data flow
subagent.completedsettled the section when the parent received a task receiptreadPromptowned inputsend()startstask_awaitreturned the task and the buffered completion wake started another turnDonebut remained mutable until another parent boundaryThe idle follower and
send()never read the parent stream concurrently. The follower preserves the active prompt key handler and draft, then hands the advanced durable cursor to the next send.Wake suppression uses framework-owned task metadata rather than matching message text. Claims are one-shot and cancellation releases them, so
input_requiredfollowed bytask_send, ordinary background completion, and cancelled awaits still wake the parent.The parent
subagent.completedevent remains a non-authoritative fallback because the final child message can still be in flight on its independent stream. Only the child stream boundary or turn cancellation finalizes the section.Regression coverage follows the complete lifecycle: background receipt, parent and user turns while the child runs, idle wake rendering without replay or draft loss, await wake coalescing, late child events staying in their section, and authoritative success or failure finalization.
Part of #1084.