[feat] Add E2B sandbox running Pi, bake harnesses into template#5045
Draft
junaway wants to merge 4 commits into
Draft
[feat] Add E2B sandbox running Pi, bake harnesses into template#5045junaway wants to merge 4 commits into
junaway wants to merge 4 commits into
Conversation
Baked E2B template (daemon+pi+node22), e2b provider branch with timeoutMs/autoPause leak backstop, refuse-restricted-network under strict, Pi asset-prep on E2B handle, E2bConfig, unit tests, docs. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…+ harnessFiles containment Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
D2: extend e2b.Dockerfile to pre-install codex, opencode, and claude (in
addition to the already-baked Pi), replicating what `install-agent` does
manually since it hangs in the E2B builder. Adds
AGENTA_AGENT_SANDBOX_{CODEX,OPENCODE,CLAUDE}_INSTALLED env levers mirroring
Pi's, documented as visibility-only since (unlike Pi) the daemon has no
skip mechanism for these three — the bake still pays off via the daemon's
own already-installed fast path.
D3: add an E2B idle-refresh keepalive (e2b-keepalive.ts) that calls the
E2B SDK's `Sandbox.setTimeout` on an interval so E2B_TIMEOUT_MS measures
idle time since the last liveness proof, matching Daytona's autostop
semantics, instead of acting as a hard wall-clock cap on the whole run.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
dosubot
Bot
added
the
size:XL
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Contributor
There was a problem hiding this comment.
Pull request overview
Adds E2B as a second remote sandbox provider for the services/runner sandbox-agent engine, including an E2B-specific relay path, idle-time keepalive semantics, workspace write hardening, and a baked E2B template image (Pi + forward-looking harnesses) to reduce cold-start latency.
Changes:
- Add E2B provider support end-to-end (plan normalization, provider creation, remote-vs-local gates, and docs/config).
- Introduce E2B timeout keepalive that refreshes
Sandbox.setTimeoutperiodically to approximate Daytona-like idle semantics. - Harden workspace file writes for both local and remote sandboxes with path-traversal containment checks, plus add/extend unit tests.
Reviewed changes
Copilot reviewed 24 out of 25 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| services/runner/src/engines/sandbox_agent/provider.ts | Adds E2B provider branch and E2B timeout/env option builder. |
| services/runner/src/engines/sandbox_agent/run-plan.ts | Introduces isE2B / isRemoteSandbox, E2B cwd + relay dir behavior, and E2B network-policy refusal. |
| services/runner/src/engines/sandbox_agent.ts | Wires E2B asset prep + E2B keepalive; generalizes remote gates from Daytona-only to any remote sandbox. |
| services/runner/src/engines/sandbox_agent/e2b.ts | Adds E2B-specific Pi asset prep and env helpers. |
| services/runner/src/engines/sandbox_agent/e2b-keepalive.ts | Implements idle-refresh keepalive via Sandbox.setTimeout. |
| services/runner/src/engines/sandbox_agent/workspace.ts | Switches to isRemoteSandbox and adds path-traversal containment for harness file writes (local + remote). |
| services/runner/src/engines/sandbox_agent/mcp.ts | Adds isRemote gating for internal MCP channel and remote relay log labeling. |
| services/runner/src/engines/sandbox_agent/usage.ts | Generalizes usage-file reads from Daytona-only to any remote sandbox. |
| services/runner/src/engines/sandbox_agent/pi-assets.ts | Adjusts local Pi asset prep gating to exclude any remote sandbox. |
| services/runner/tests/unit/session-mcp-layering.test.ts | Updates MCP layering tests for new isRemote input. |
| services/runner/tests/unit/sandbox-agent-workspace.test.ts | Updates workspace tests for isRemoteSandbox and adds E2B + path-traversal test coverage. |
| services/runner/tests/unit/sandbox-agent-usage.test.ts | Updates usage tests for isRemote flag. |
| services/runner/tests/unit/sandbox-agent-run-plan.test.ts | Extends existing run-plan tests to assert isRemoteSandbox. |
| services/runner/tests/unit/sandbox-agent-e2b-run-plan.test.ts | New unit tests for E2B run-plan behavior and E2B network-policy refusal. |
| services/runner/tests/unit/sandbox-agent-e2b-provider.test.ts | New unit tests for E2B create-option construction + timeout leak backstop behavior. |
| services/runner/tests/unit/sandbox-agent-e2b-keepalive.test.ts | New unit tests for keepalive scheduling, stop semantics, and error swallowing. |
| services/runner/sandbox-images/e2b/e2b.Dockerfile | New baked E2B template that preinstalls sandbox-agent + harnesses (incl. Node 22). |
| services/runner/sandbox-images/e2b/README.md | Documentation for building/configuring the baked E2B template and timeout semantics. |
| services/runner/package.json | Adds @e2b/code-interpreter dependency. |
| services/runner/pnpm-lock.yaml | Locks @e2b/code-interpreter and transitive deps. |
| docs/design/agent-workflows/projects/add-sandbox-e2b/tasks.md | Project task breakdown for E2B sandbox addition. |
| docs/design/agent-workflows/projects/add-sandbox-e2b/specs.md | Design/specs and acceptance criteria for E2B sandbox support. |
| docs/design/agent-workflows/projects/add-sandbox-e2b/research.md | Research notes and provider constraints (network policy, template build gotchas, etc.). |
| docs/design/agent-workflows/documentation/running-the-agent.md | Updates env var docs to include e2b provider + E2B env vars. |
| api/oss/src/utils/env.py | Adds E2BConfig to the shared env object. |
Files not reviewed (1)
- services/runner/pnpm-lock.yaml: Generated file
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+141
to
+147
| envs: { | ||
| AGENTA_AGENT_SANDBOX_CODEX_INSTALLED: String(E2B_CODEX_INSTALLED), | ||
| AGENTA_AGENT_SANDBOX_OPENCODE_INSTALLED: String(E2B_OPENCODE_INSTALLED), | ||
| AGENTA_AGENT_SANDBOX_CLAUDE_INSTALLED: String(E2B_CLAUDE_INSTALLED), | ||
| ...piExtEnv, | ||
| ...secrets, | ||
| }, |
This was referenced Jul 4, 2026
Merged
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
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
The only remote sandbox supported so far was Daytona. E2B is a second remote provider
sandbox-agentalready exports support for; standing it up meant a new provider branch, a new relay path (E2B's sandbox FS API instead of a geesefs-mounted cwd), and a new keepalive story since E2B has no built-in idle timeout like Daytona.Changes
provider.ts: E2B branch alongside Daytona, with a typedE2BCreateOptionsreturn (dropped anas anycast).run-plan.ts: newisRemoteSandboxflag derived asisDaytona || isE2B, used at every pure-remoteness gate; identity-specific sites (cookie fetch, cwd factory, network policy, MCP relay log label) keep the specificisE2B/isDaytonaflags. E2B's relay dir is${cwd}/.agenta-tools, polled via the sandbox FS API — the E2B cwd is never geesefs-mounted, unlike Daytona/local.workspace.ts: gates onisRemoteSandbox(was Daytona-only) so E2B also gets sandbox-FS-backed harness file writes, plus path-traversal containment (assertContainedfor host paths,assertContainedPosixfor remote) before every write.e2b-keepalive.ts(new): since E2B has no built-in idle-based timeout like Daytona,startE2BKeepaliverefreshesSandbox.setTimeouteverytimeoutMs/3for the life of the session, giving it Daytona-like idle semantics instead of a hard wall-clock cutoff.Tests / notes
sandbox-agent-e2b-keepalive.test.ts,sandbox-agent-e2b-provider.test.ts,sandbox-agent-e2b-run-plan.test.ts, plus updates tosandbox-agent-workspace.test.tsandsession-mcp-layering.test.ts.chore/add-claude-e2b(its PR targets this branch, notbig-agents).chore/add-codex-e2bandchore/add-opencode-e2breuse this E2B plumbing but are stacked on their respective harness branches instead.