Skip to content

fix(eve): count only new input toward the session input-token limit - #5011

Draft
ruiconti wants to merge 2 commits into
mainfrom
budget-count-new-input
Draft

ruiconti wants to merge 2 commits into
mainfrom
budget-count-new-input

Conversation

@ruiconti

Copy link
Copy Markdown
Contributor

Summary

The session input-token limit counted cache reads at full weight. A long, healthy, tool-heavy session replays its history from the prompt cache on every step, so almost all of its input is cache reads, and it hit the continuation prompt again and again with no loop in sight. The limit now counts only new input: uncached input plus cache writes (inputTokens - cacheReadTokens). This applies to the limit check, the window granted on Continue, the remaining quota, and the shares that subagents inherit. Usage telemetry and event usage still report raw input and cache totals. A provider that reports no cache details counts exactly as before. The 40_000_000 default stays.

Counting cache writes alone was not an option: providers with automatic caching report no cache writes, so the limit would never fire for them.

This is the first of a series. Follow-ups will make one approval renew the whole delegation tree and add structured data to the session-limit request.

Validation

  • vitest.unit.config.ts on src/harness, src/subagents, src/execution/tasks, src/execution/session: 85 files, 1274 tests pass. New tests in turn-tag-state.test.ts run a model step through extractTokenUsageDelta with inputTokenDetails set (cache reads excluded from the limit, grant and remaining quota; raw totals kept) and with it missing (full input counts, as before). The input case of the fail-fast test in tool-loop.test.ts now has 14 raw input with 2 cache reads, so it reaches the 12-token limit and reports the raw total in details.
  • vitest.integration.config.ts on the session-limit cancellation, session entry and approval-resume suites: 4 files pass.
  • pnpm --filter eve typecheck, pnpm lint, pnpm guard:invariants pass.

Checklist

  • 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)

Signed-off-by: ruiconti <1834568+ruiconti@users.noreply.github.com>
@vercel

vercel Bot commented Oct 19, 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 Oct 19, 2026 3:16am UTC
eve-pkg Ready Ready Preview, v0 Oct 19, 2026 3:16am UTC

Comment thread docs/agent-config.md

When `maxInputTokensPerSession` is omitted, root sessions apply a default
input budget of `40_000_000` provider-reported input tokens.
input budget of `40_000_000` new input tokens.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The docs now say the default is 40M new input tokens, but the authored API contract still says the opposite. The maxInputTokensPerSession JSDoc in packages/eve/src/shared/agent-definition.ts:188 (what authors see on hover in agent.ts) still reads "Maximum provider-reported input tokens accumulated by one durable session", and SessionLimits.maxInputTokensPerSession in packages/eve/src/harness/types.ts:121 says the same. After this PR the limit is inputTokens - cacheReadTokens (getSessionBudgetUsage), so both comments describe the wrong quantity. Can we update them to match: new input (uncached plus cache writes), with cache reads excluded, and full input counted when the provider reports no cache details?

@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

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

Key takeaways

  • No notable deltas vs main (c1a6fb1).

Delta vs main (c1a6fb1)

Area Metric Baseline Current Delta
Package Packed tarball 7.66 MB 7.67 MB +219 B ⚠️
Package Unpacked publish size 26.98 MB 26.98 MB +457 B ⚠️
Package Installed footprint 74.94 MB 74.94 MB +457 B ⚠️
Package Published files 4171 4171 0
Package Installed files 8150 8150 0
Package Installed package instances 34 34 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 20.78 MB 20.78 MB +1.9 kB ⚠️
Runtime Public routes 18 18 0
Changed function payloads vs main (c1a6fb1) (2)
Function Status Baseline Current Delta Route changes
functions/__server.func changed 10.39 MB 10.39 MB +936 B ⚠️ none
functions/.well-known/workflow/v1/flow.func changed 10.39 MB 10.39 MB +936 B ⚠️ none

eve init install

Metric Baseline Current Delta
Installed footprint 113.32 MB 113.32 MB +457 B ⚠️
Installed packages 97 97 0
dependencies 4 4 0
devDependencies 2 2 0
Dependency package bytes 45.26 MB 45.26 MB +457 B ⚠️
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: 2026-10-19T03:17:25.158Z
  • Route aliases: 18 public, 1 internal (19 total aliases)
  • Vercel routes in config: 21
  • Severity legend: 🔴 dominant/large, 🟠 notable, 🟡 watch, ⚪ small
Package Drill-Down

Package Details

  • Package: eve@0.80.1
  • Package directory: packages/eve
  • Tarball: 7.67 MB (eve-0.80.1.tgz)
  • Unpacked payload: 26.98 MB across 4171 published files
  • Installed footprint: 74.94 MB across 8150 installed files
  • Installed root package: 26.98 MB
  • Installed dependencies: 47.96 MB
  • Installed package instances: 34
  • 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: 26.98 MB (36.0%)
  • @rolldown/binding-linux-x64-gnu: 19.81 MB (26.4%)
  • ai: 8.24 MB (11.0%)
  • zod: 6.15 MB (8.2%)
  • undici: 3.58 MB (4.8%)
Publish payload breakdown
Published file size
🟠 dist/src/compiled/shadcn-registry/index.js       [#####...................] 3.85 MB 14.3%
🟠 dist/src/compiled/@photon-ai/chat-adapter-ime... [###.....................] 2.29 MB 8.5%
🟠 dist/src/compiled/@ai-sdk/code-mode/index.js     [#.......................] 1.03 MB 3.8%
🟡 dist/src/compiled/@vercel/blob/index.js          [#.......................] 604.9 kB 2.2%
🟡 dist/src/compiled/_chunks/workflow/undici-D-I... [#.......................] 521.9 kB 1.9%
🔴 Other published files                            [########################] 18.69 MB 69.3%
Installed footprint breakdown
Installed package size
🔴 eve                             [########################] 26.98 MB 36.0%
🔴 @rolldown/binding-linux-x64-gnu [##################......] 19.81 MB 26.4%
🔴 ai                              [#######.................] 8.24 MB 11.0%
🔴 zod                             [#####...................] 6.15 MB 8.2%
🟠 undici                          [###.....................] 3.58 MB 4.8%
🟠 nitro                           [##......................] 1.89 MB 2.5%
🔴 Other installed packages        [#######.................] 8.28 MB 11.0%
Runtime dependencies (2)
Package Range Notes
nitro 3.0.260903-beta
undici 8.10.2
Peer dependencies (7)
Package Range Notes
@opentelemetry/api ^1.0.0 optional peer
ai catalog:
braintrust ^3.0.0 optional peer
chat ^4.42.0 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: 113.32 MB across 10067 installed files
  • Installed packages: 97 total (91 transitive-only)
  • dependencies: 4 direct packages totaling 45.26 MB
  • devDependencies: 2 direct packages totaling 5.11 MB
  • Other transitive package files: 62.95 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 (24.7%)
  • eve: 26.98 MB (23.8%)
  • @rolldown/binding-linux-x64-gnu: 19.81 MB (17.5%)
  • zod: 9.76 MB (8.6%)
  • ai: 8.24 MB (7.3%)
Installed footprint breakdown
Installed package size
🔴 @typescript/typescript-linux-x64 [########################] 27.95 MB 24.7%
🔴 eve                              [#######################.] 26.98 MB 23.8%
🔴 @rolldown/binding-linux-x64-gnu  [#################.......] 19.81 MB 17.5%
🔴 zod                              [########................] 9.76 MB 8.6%
🔴 ai                               [#######.................] 8.24 MB 7.3%
🟠 undici                           [###.....................] 3.58 MB 3.2%
🔴 Other installed packages         [###############.........] 17.00 MB 15.0%
dependencies (4)
Package Range Installed size Share
@vercel/connect 2.5.2 276.3 kB 0.2%
ai ^7.0.137 8.24 MB 7.3%
eve file:eve-0.80.1.tgz 26.98 MB 23.8%
zod 4.5.4 9.76 MB 8.6%
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     [########################] 10.39 MB 50.0%
🔴 functions/__server.func                         [########################] 10.39 MB 50.0%

Top Function Payloads

🟠 functions/.well-known/workflow/v1/flow.func • 1 public route • 10.39 MB
Metric Value
Public routes /.well-known/workflow/v1/flow
Runtime nodejs24.x
Handler index.mjs
Payload 10.39 MB
Function files 10.39 MB across 122 files
Traced dependencies 0 B
Signal 🟠 Bundled file _chunks/vercel.web.mjs is 2.38 MB (23.0%)

🟠 🔎 Dependency Analysis

📦 Bundled files:

Bundled file size
🟠 _chunks/vercel.web.mjs                         [############............] 2.38 MB 23.0%
🟠 _libs/undici.mjs                               [#####...................] 1.01 MB 9.7%
🟡 _chunks/sandbox3.mjs                           [####....................] 815.4 kB 7.8%
🟡 _chunks/compiled-artifacts-instrumentation.mjs [####....................] 734.8 kB 7.1%
🟡 _chunks/token-util-Bla5Z0G9.mjs                [####....................] 704.7 kB 6.8%
🔴 Other bundled files                            [########################] 4.74 MB 45.7%

🧾 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 • 17 public routes, 1 internal alias • 10.39 MB
Metric Value
Public routes /
/.well-known/workflow/v1/webhook/[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
/eve/v1/session/[sessionId]/stubs
/eve/v1/task-input/[token]
Internal aliases /__server
Runtime nodejs24.x
Handler index.mjs
Payload 10.39 MB
Function files 10.39 MB across 122 files
Traced dependencies 0 B
Signal 🟠 Bundled file _chunks/vercel.web.mjs is 2.38 MB (23.0%)

🟠 🔎 Dependency Analysis

📦 Bundled files:

Bundled file size
🟠 _chunks/vercel.web.mjs                         [############............] 2.38 MB 23.0%
🟠 _libs/undici.mjs                               [#####...................] 1.01 MB 9.7%
🟡 _chunks/sandbox3.mjs                           [####....................] 815.4 kB 7.8%
🟡 _chunks/compiled-artifacts-instrumentation.mjs [####....................] 734.8 kB 7.1%
🟡 _chunks/token-util-Bla5Z0G9.mjs                [####....................] 704.7 kB 6.8%
🔴 Other bundled files                            [########################] 4.74 MB 45.7%

🧾 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: 3.28 s -> 3.20 s (-85.6 ms) vs main (c1a6fb1).
  • Timing is informational: shared GitHub runners are too variable for a hard timing budget.
Detailed phase timings vs `main (c1a6fb1)`
Phase Baseline Current Delta
extension.check 0.4 ms 0.3 ms -0.1 ms
project.resolve 0.3 ms 0.2 ms -0.1 ms
workspace.create 0.4 ms 0.4 ms 0.0 ms
host.prepare 779.9 ms 729.9 ms -50.0 ms
vercel.service-prefix.resolve 1.3 ms 1.2 ms -0.1 ms
nitro.create 517.0 ms 503.4 ms -13.6 ms
sandbox.prewarm 425.6 ms 428.3 ms +2.7 ms
nitro.cache.prepare 0.2 ms 0.2 ms 0.0 ms
nitro.prepare 0.6 ms 0.6 ms 0.0 ms
nitro.public-assets 0.6 ms 0.7 ms +0.1 ms
nitro.prerender 0.4 ms 0.4 ms 0.0 ms
nitro.bundle 1.42 s 1.40 s -23.5 ms
nitro.cache.write 0.3 ms 0.2 ms -0.1 ms
vercel.workflow-function.materialize 69.7 ms 69.8 ms +0.1 ms
agent-summary.emit 0.6 ms 0.6 ms 0.0 ms
connect-manifest.emit 0.3 ms 0.2 ms -0.1 ms
nitro.close 0.1 ms 0.1 ms 0.0 ms
output.publish 2.9 ms 3.0 ms +0.1 ms
workspace.remove 2.1 ms 2.1 ms 0.0 ms

Signed-off-by: ruiconti <1834568+ruiconti@users.noreply.github.com>

@vercel vercel Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Additional Suggestion:

JSDoc for maxInputTokensPerSession still describes "provider-reported input tokens accumulated" even though the limit now counts only new input (inputTokens - cacheReadTokens), making the public-API hover documentation inaccurate.

Fix on Vercel

This branch was successfully deployed

2 active deployments
Preview – eve-docs — 439b5c1b Deployed Oct 19, 2026 by vercel[bot]
Preview – eve-pkg — 439b5c1b Deployed Oct 19, 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