Skip to content

refactor(tui): run eve dev on the shared agent store - #3880

Open
OwenKephart wants to merge 8 commits into
refactor/shared-conversation-clientfrom
refactor/tui-agent-store
Open

OwenKephart wants to merge 8 commits into
refactor/shared-conversation-clientfrom
refactor/tui-agent-store

Conversation

@OwenKephart

@OwenKephart OwenKephart commented Sep 27, 2026 •

Copy link
Copy Markdown
Contributor

Summary

The eve dev TUI ran its own steering stream, subagent pump, message projection, and turn ledger. It duplicated the lifecycle the web store already owns and kept drifting from web behavior. It now consumes EveAgentStore like the framework hooks. It projects its transcript from store snapshots and keeps one persistent composer. The prompt stays open while the agent works, including turns the agent starts on its own. Slash commands run immediately. Approvals and questions open as soon as they arrive, and Esc or Ctrl+C cancels the running turn.

Each agent call gets its own section, which stays live until its reply arrives, and a section keeps its place above tool rows that appeared before its agent started. Tool rows are grouped by step and stay running until their task settles. The activity line names the tasks a held turn is waiting on. Each answer to an approval or question is sent as soon as it is given, and when several are open the prompt shows its place among them and the subagent or task that asked.

Removed behavior: the mid-turn message queue (and steering with queued messages), restoring the prompt after a cancel, the connection-authorization status label, and "preparing" placeholders for tool rows.

Third of four stacked PRs, on top of #3879. #3965 keeps a task call's tool part running until its task settles, which replaces this PR's working-task check.

Bugs fixed

  • When one model call finished some text, called a tool, then wrote more text under the same step index, the second message was dropped.
  • A late step.completed for one step also closed a later step's still-streaming text and reasoning.
  • In full reasoning mode, the final reasoning didn't replace the streamed draft when they differed.
  • Approvals and questions that settled later in the same stream were still prompted.
  • Answered requests were tracked per stream, so a redelivered event on a later stream asked the same approval again.
  • Duplicate events were only filtered within each stream, so reconnects could duplicate content.
  • A connection-authorization completion updated the row of another attempt with the same connection name.
  • Pressing Esc before the turn ID arrived sent an unguarded cancel and retried it for two seconds, which could cancel a later turn.
  • Failed or cancelled subagents showed as "Done".
  • When a model step asked for an approval and also started a subagent, the approved tool's row still showed it awaiting approval after it ran.
  • While an approval was open, a turn the agent started on its own, such as a scheduled run, didn't render until you answered.

Validation

  • pnpm typecheck, pnpm lint, pnpm fmt, pnpm guard:invariants, pnpm docs:check, git diff --check
  • Unit suite: 799 files, 8,524 passed, 1 skipped
  • Integration suite: 129 files, 987 passed, including runner tests against an in-memory eve server
  • After rebasing onto refactor(client): share one conversation client across the store and hooks #3879's kind change: tsc and unit tests for src/cli/dev/tui and src/client (66 files, 1,140 passed), and their integration tests (6 files, 28 passed)
  • All 14 TUI smoke scripts (node test/tui-client/run-all.mjs) against mock-model and live-model fixtures
  • Manual, driving the TUI through its test runner against mock-model agents on both the tasks base and this branch: the approved tool row, the cancelled subagent, and the failed subagent from the list above.

Checklist

  • This change was requested or approved by a maintainer
  • I ran the relevant checks from CONTRIBUTING.md
  • I added tests and documentation where relevant
  • I added a changeset if this touches the published eve package
  • DCO sign-off passes for every commit (git commit --signoff)

@OwenKephart
OwenKephart added this pull request to stack #3881 September 27, 2026 00:54
@vercel

vercel Bot commented Sep 27, 2026 •

Copy link
Copy Markdown
Contributor

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

Project Deployment Actions Updated
eve-docs Ready Ready Preview, v0 Sep 30, 2026 3:11am UTC
eve-pkg Ready Ready Preview, v0 Sep 30, 2026 3:11am UTC

@github-actions

github-actions Bot commented Sep 27, 2026 •

Copy link
Copy Markdown
Contributor

Bundle + Package Summary: apps/fixtures/weather-agent

Key takeaways

  • No notable deltas vs refactor/shared-conversation-client (3e63c13).

Delta vs refactor/shared-conversation-client (3e63c13)

Area Metric Baseline Current Delta
Package Packed tarball 8.45 MB 8.45 MB -4.4 kB ✅
Package Unpacked publish size 31.04 MB 31.01 MB -21.1 kB ✅
Package Installed footprint 77.43 MB 77.40 MB -21.1 kB ✅
Package Published files 3908 3914 +6
Package Installed files 7887 7893 +6
Package Installed package instances 33 33 0
Package Distinct installed package names 32 32 0
Package Installed dependency edges 51 51 0
Package Installed optional peer edges 9 9 0
Runtime Unique function payloads 2 2 0
Runtime Total function bytes 19.08 MB 19.08 MB -80 B ✅
Runtime Public routes 17 17 0
Changed function payloads vs refactor/shared-conversation-client (3e63c13) (2)
Function Status Baseline Current Delta Route changes
functions/__server.func changed 9.54 MB 9.54 MB -40 B ✅ none
functions/.well-known/workflow/v1/flow.func changed 9.54 MB 9.54 MB -40 B ✅ none

eve init install

Metric Baseline Current Delta
Installed footprint 115.72 MB 115.70 MB -21.1 kB ✅
Installed packages 97 97 0
dependencies 4 4 0
devDependencies 2 2 0
Dependency package bytes 48.41 MB 48.38 MB -21.1 kB ✅
devDependency package bytes 5.11 MB 5.11 MB 0 B ➖
Build Metadata
  • Preset: vercel
  • Nitro: nitro@3.0.260903-beta
  • Output directory: apps/fixtures/weather-agent/.vercel/output
  • Build metadata timestamp: 2026-09-30T03:12:20.334Z
  • Route aliases: 17 public, 1 internal (18 total aliases)
  • Vercel routes in config: 20
  • Severity legend: 🔴 dominant/large, 🟠 notable, 🟡 watch, ⚪ small
Package Drill-Down

Package Details

  • Package: eve@0.68.0
  • Package directory: packages/eve
  • Tarball: 8.45 MB (eve-0.68.0.tgz)
  • Unpacked payload: 31.01 MB across 3914 published files
  • Installed footprint: 77.40 MB across 7893 installed files
  • Installed root package: 30.73 MB
  • Installed dependencies: 46.68 MB
  • Installed package instances: 33
  • Distinct installed package names: 32
  • Installed dependency edges: 51
  • Installed optional peer edges: 9
  • Runtime dependencies: 2
  • Peer dependencies: 7 (6 optional)

Installed footprint is measured from an isolated temporary npm install of the packed tarball.
Graph metrics read only package.json files in package directories directly beneath a node_modules boundary, including nested boundaries. Each directory is one package instance; distinct names come from those manifests. Dependency edges count each unique name in dependencies or optionalDependencies per instance; optional peer edges count peerDependencies marked optional.

Heavy installed dependencies

  • eve: 30.73 MB (39.7%)
  • @rolldown/binding-linux-x64-gnu: 19.15 MB (24.7%)
  • ai: 7.71 MB (10.0%)
  • zod: 6.14 MB (7.9%)
  • undici: 3.52 MB (4.5%)
Publish payload breakdown
Published file size
🔴 dist/src/compiled/shadcn-registry/index.js       [##############..........] 9.70 MB 31.3%
🟠 dist/src/compiled/@photon-ai/chat-adapter-ime... [###.....................] 2.29 MB 7.4%
🟠 dist/src/compiled/@ai-sdk/code-mode/index.js     [#.......................] 1.03 MB 3.3%
🟡 dist/src/compiled/@vercel/blob/index.js          [#.......................] 604.9 kB 2.0%
🟡 dist/src/compiled/_chunks/workflow/signal-exi... [#.......................] 514.5 kB 1.7%
🔴 Other published files                            [########################] 16.88 MB 54.4%
Installed footprint breakdown
Installed package size
🔴 eve                             [########################] 30.73 MB 39.7%
🔴 @rolldown/binding-linux-x64-gnu [###############.........] 19.15 MB 24.7%
🔴 ai                              [######..................] 7.71 MB 10.0%
🔴 zod                             [#####...................] 6.14 MB 7.9%
🟠 undici                          [###.....................] 3.52 MB 4.5%
🟠 nitro                           [#.......................] 1.89 MB 2.4%
🔴 Other installed packages        [######..................] 8.27 MB 10.7%
Runtime dependencies (2)
Package Range Notes
nitro 3.0.260903-beta
undici 8.9.0
Peer dependencies (7)
Package Range Notes
@opentelemetry/api ^1.0.0 optional peer
ai catalog:
braintrust ^3.0.0 optional peer
chat ^4.41.0 optional peer
dd-trace ^6.13.0 optional peer
just-bash ^3.1.0 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: 115.70 MB across 9788 installed files
  • Installed packages: 97 total (91 transitive-only)
  • dependencies: 4 direct packages totaling 48.38 MB
  • devDependencies: 2 direct packages totaling 5.11 MB
  • Other transitive package files: 62.20 MB

Installed footprint is measured from an isolated temporary eve init my-agent using the current packed eve tarball.

Heavy installed dependencies

  • eve: 30.73 MB (26.6%)
  • @typescript/typescript-linux-x64: 27.95 MB (24.2%)
  • @rolldown/binding-linux-x64-gnu: 19.15 MB (16.5%)
  • zod: 9.76 MB (8.4%)
  • ai: 7.71 MB (6.7%)
Installed footprint breakdown
Installed package size
🔴 eve                              [########################] 30.73 MB 26.6%
🔴 @typescript/typescript-linux-x64 [######################..] 27.95 MB 24.2%
🔴 @rolldown/binding-linux-x64-gnu  [###############.........] 19.15 MB 16.5%
🔴 zod                              [########................] 9.76 MB 8.4%
🔴 ai                               [######..................] 7.71 MB 6.7%
🟠 undici                           [###.....................] 3.52 MB 3.0%
🔴 Other installed packages         [#############...........] 16.90 MB 14.6%
dependencies (4)
Package Range Installed size Share
@vercel/connect 2.2.0 194.7 kB 0.2%
ai ^7.0.105 7.71 MB 6.7%
eve file:eve-0.68.0.tgz 30.73 MB 26.6%
zod 4.5.4 9.76 MB 8.4%
devDependencies (2)
Package Range Installed size Share
@types/node 24.x 2.61 MB 2.3%
typescript 7.0.2 2.50 MB 2.2%
Function Drill-Down

Payload Size Graph

Unique function payload size and share of total
🔴 functions/.well-known/workflow/v1/flow.func     [########################] 9.54 MB 50.0%
🔴 functions/__server.func                         [########################] 9.54 MB 50.0%

Top Function Payloads

🟠 functions/.well-known/workflow/v1/flow.func • 1 public route • 9.54 MB
Metric Value
Public routes /.well-known/workflow/v1/flow
Runtime nodejs24.x
Handler index.mjs
Payload 9.54 MB
Function files 9.54 MB across 111 files
Traced dependencies 0 B
Signal 🟠 Bundled file _chunks/vercel.web.mjs is 2.15 MB (22.5%)

🟠 🔎 Dependency Analysis

📦 Bundled files:

Bundled file size
🟠 _chunks/vercel.web.mjs                         [############............] 2.15 MB 22.5%
🟡 _libs/undici.mjs                               [######..................] 980.8 kB 10.3%
🟡 _chunks/sandbox.mjs                            [#####...................] 811.4 kB 8.5%
🟡 _chunks/compiled-artifacts-instrumentation.mjs [####....................] 756.3 kB 7.9%
🟡 _chunks/signal-exit-DZKTacNU.mjs               [###.....................] 616.2 kB 6.5%
🔴 Other bundled files                            [########################] 4.23 MB 44.3%

🧾 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 • 16 public routes, 1 internal alias • 9.54 MB
Metric Value
Public routes /
/.well-known/workflow/v1/webhook/[token]
/eve/v1/activity/[token]
/eve/v1/callback/[token]
/eve/v1/connections/[name]/callback/[attemptId]/[token]
/eve/v1/connections/[name]/callback/[token]
/eve/v1/health
/eve/v1/info
/eve/v1/session
/eve/v1/session/[parentSessionId]/subagents/[callId]/[childSessionId]/stream
/eve/v1/session/[sessionId]
/eve/v1/session/[sessionId]/cancel
/eve/v1/session/[sessionId]/clear
/eve/v1/session/[sessionId]/compact
/eve/v1/session/[sessionId]/reset
/eve/v1/session/[sessionId]/stream
Internal aliases /__server
Runtime nodejs24.x
Handler index.mjs
Payload 9.54 MB
Function files 9.54 MB across 111 files
Traced dependencies 0 B
Signal 🟠 Bundled file _chunks/vercel.web.mjs is 2.15 MB (22.5%)

🟠 🔎 Dependency Analysis

📦 Bundled files:

Bundled file size
🟠 _chunks/vercel.web.mjs                         [############............] 2.15 MB 22.5%
🟡 _libs/undici.mjs                               [######..................] 980.8 kB 10.3%
🟡 _chunks/sandbox.mjs                            [#####...................] 811.4 kB 8.5%
🟡 _chunks/compiled-artifacts-instrumentation.mjs [####....................] 756.3 kB 7.9%
🟡 _chunks/signal-exit-DZKTacNU.mjs               [###.....................] 616.2 kB 6.5%
🔴 Other bundled files                            [########################] 4.23 MB 44.3%

🧾 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: 4.39 s -> 4.41 s (+23.1 ms) vs refactor/shared-conversation-client (3e63c13).
  • Timing is informational: shared GitHub runners are too variable for a hard timing budget.
Detailed phase timings vs `refactor/shared-conversation-client (3e63c13)`
Phase Baseline Current Delta
extension.check 12.3 ms 4.1 ms -8.2 ms
project.resolve 1.0 ms 2.2 ms +1.2 ms
workspace.create 1.0 ms 0.9 ms -0.1 ms
host.prepare 806.9 ms 929.9 ms +123.0 ms
vercel.service-prefix.resolve 1.9 ms 1.8 ms -0.1 ms
nitro.create 615.6 ms 620.2 ms +4.6 ms
sandbox.prewarm 337.1 ms 303.6 ms -33.5 ms
nitro.cache.prepare 0.2 ms 0.3 ms +0.1 ms
nitro.prepare 0.7 ms 0.8 ms +0.1 ms
nitro.public-assets 0.8 ms 0.8 ms 0.0 ms
nitro.prerender 0.5 ms 0.5 ms 0.0 ms
nitro.bundle 2.46 s 2.40 s -58.7 ms
nitro.cache.write 0.5 ms 0.5 ms 0.0 ms
vercel.workflow-function.materialize 54.2 ms 54.5 ms +0.3 ms
agent-summary.emit 0.7 ms 0.7 ms 0.0 ms
connect-manifest.emit 0.3 ms 0.3 ms 0.0 ms
nitro.close 0.1 ms 0.1 ms 0.0 ms
output.publish 3.6 ms 3.7 ms +0.1 ms
workspace.remove 2.8 ms 2.8 ms 0.0 ms

Comment thread packages/eve/src/cli/dev/tui/runner.ts
Comment thread packages/eve/src/cli/dev/tui/terminal-renderer.ts Outdated
Comment thread packages/eve/src/cli/dev/tui/errors.ts Outdated
Comment thread packages/eve/src/cli/dev/tui/transcript.ts Outdated
Comment thread packages/eve/src/cli/dev/tui/runner.ts Outdated
Comment thread packages/eve/src/cli/dev/tui/errors.ts Outdated
Comment thread packages/eve/src/cli/dev/tui/errors.ts Outdated
Comment thread packages/eve/src/cli/dev/tui/transcript.ts Outdated
Comment thread packages/eve/src/cli/dev/tui/runner.ts
Signed-off-by: owenkephart <owen.kephart@vercel.com>
Signed-off-by: owenkephart <owen.kephart@vercel.com>
Signed-off-by: owenkephart <owen.kephart@vercel.com>
The requester and the "2 of 10" position took a line above the prompt's title. They now sit together at the right end of the drawer's top rule, as "subagent(color_generator) · 1 of 4", clipped with an ellipsis when the terminal is too narrow.

Signed-off-by: owenkephart <owen.kephart@vercel.com>
A root approval ends the turn even while a task from the same step keeps working. The transcript treated every running row in an ended turn as interrupted, so an agent call showed "interrupted" while its approval drawer was open. A working task call now stays running until its task.settled.

Signed-off-by: owenkephart <owen.kephart@vercel.com>
Signed-off-by: owenkephart <owen.kephart@vercel.com>
Whether a call's content can still arrive relies on the same rule as `agentCallTurns` and `isAgentSessionCaughtUp`: the session's k-th message came from the task's k-th call. The transcript kept its own copy of that check. It now lives with the other two as `isAgentCallContentPending`, which `eve/client` still does not export.

Signed-off-by: owenkephart <owen.kephart@vercel.com>
…store

main now writes a task's start and end as single lines and shows what
working tasks do in a panel above the prompt (#3957), which replaced the
live subagent sections this stack had kept. The store-based transcript
now renders the same display: a task call becomes its start line; end
lines and --subagents full rows are placed where the transcript had
reached when they appeared; the panel, the Waiting for turn bar, stopped
tasks, hidden task tools, preparing placeholders, and root tool labels
and refusals match main. An approval or question names the task that
asked as its lines do.

Signed-off-by: owenkephart <owen.kephart@vercel.com>

This branch was successfully deployed

2 active deployments
Preview – eve-pkg — 8653e02d Deployed Sep 30, 2026 by vercel[bot]
Preview – eve-docs — 8653e02d Deployed Sep 30, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant