Releases: jinzijian/EvoTrace
Release list
EvoTrace v0.3.2 — one-command onboarding
One-command local onboarding for Claude Code and Codex history.
Highlights:
evotrace initdiscovers, imports, and mines existing sessions- One-line installers for macOS, Linux, WSL, and Windows PowerShell
- Value-first GitHub landing page and terminal product visual
- Windows-safe CLI imports with a real PowerShell installer smoke test
- Bilingual documentation and GitHub Discussions
EvoTrace remains local-first: installation reads no session data, and init performs no model call or upload.
EvoTrace v0.3.1 — one-command onboarding
EvoTrace can now go from installation to a locally mined asset index with one command.
Highlights:
evotrace initdiscovers, imports, and mines Claude Code and Codex history- One-line installers for macOS, Linux, WSL, and Windows PowerShell
- Unsloth-inspired GitHub landing page with a value-first quickstart
- New terminal product visual, bilingual docs, and community links
- Installer and onboarding regression coverage
EvoTrace remains local-first: installation reads no session data, and init performs no model call or upload.
EvoTrace v0.3.0 — project rename
ScaleVerifier is now EvoTrace.
Turn every Claude Code and Codex session into reusable training, evaluation, and verification assets.
What changed
- The GitHub repository is now
jinzijian/evotrace. - The Python distribution and primary import package are now
evotrace. - The primary CLI is
evotrace;etis the short alias. - The default local store is
~/.evotrace, configurable withEVOTRACE_HOME. - Generated bundle metadata, Docker paths, environment variables, Git identity, documentation, and security language now use EvoTrace.
Compatibility
- Existing
~/.scaleverifierdata is reused automatically when no new store exists. scaleverifier,vf, andsvCLI aliases remain available.- The
scaleverifierPython namespace remains as a compatibility shim. - Legacy
SCALEVERIFIER_HOMEand task environment variables remain supported. - GitHub redirects the former repository URL to the new one.
Verification
10 local tests pass, including brand migration and backward-compatibility coverage. CI passes install, test, lint, and build on Python 3.9, 3.11, and 3.13.
ScaleVerifier v0.2.0 — import, mine, build
ScaleVerifier v0.2.0 turns existing Claude Code and Codex history into a day-one local eval pipeline.
Highlights
vf importdiscovers both local history sources, distinguishes rich Codex sessions from prompt history, and incrementally indexes changed files.vf mineproduces auditable useful, correction, execution, preference, recovery, and low-value labels without calling a model.vf buildcompiles top execution-verifiable sessions into task, Git environment, Dockerfile, and verifier bundles.vf watchkeeps the index and mining results current for nightly workflows.- Every bundle includes
sandbox-policy.jsonand a non-root container image contract.
Safety boundary
Autonomous agents may work only on the ephemeral container copy. Host bind mounts, Docker socket access, privileged mode, production write credentials, and host-side agent execution are outside the contract. V0.2 rejects legacy benchmark --agent; a safe container orchestrator is the next runtime milestone.
Behavior changes
- The default local store is now
~/.scaleverifier(override withSCALEVERIFIER_HOMEor--home). vfis the primary CLI;scaleverifierandsvremain aliases.
Verification
CI passes on Python 3.9, 3.11, and 3.13. The generated bundle Dockerfile was also built and launched under non-root, network-disabled, capability-dropped smoke-test settings.
ScaleVerifier v0.1.0
First public alpha of ScaleVerifier.
Highlights:
- Import local Codex and Claude Code histories into a normalized trajectory schema
- Record arbitrary coding-agent commands with Git before/after state
- Compile sessions into portable task, environment, and verifier bundles
- Replay tasks and benchmark agents in fresh workspaces
- Mine basic failure signals locally
- Local-first storage, best-effort secret redaction, and explicit privacy boundaries
ScaleVerifier is early alpha. Inspect generated verifiers and bundles before consequential use or sharing.