Specula v1.1.0

Specula v1.1.0 makes long-running investigations easier to resume, their results easier to review, and target-specific modeling priorities easier to carry through the workflow. It also improves --keep-original patches and refreshes the curated case-study corpus.

Highlights

• specula run --run-id now resumes the exact unfinished agent conversation across Claude Code, Codex, Copilot CLI, OpenCode, and Pi. Specula restores the run settings, returns to the interrupted phase, rejects live or mismatched session owners, and stops rather than silently starting a new conversation when resumable state is unavailable. Use --fresh-context to deliberately continue from the retained files with a new agent context.
• Each target now has a primary human-readable .specula-output/summary.md. It reports completion status, confirmed and limited findings, validation limits, source and attempt commits, links to detailed evidence, and deterministic resource information without surfacing stale or invalid results.
• Resource summaries record per-phase duration, total and cached tokens, adapter-reported estimated cost, configured parallelism, and TLC limits. Measurements accumulate across resumes, unavailable values remain explicit, and partial totals are marked incomplete.
• Single-target runs can use --guidance=PATH to provide target-specific goals, scope, priority questions, interactions, and assumptions. The guidance is preserved through code analysis and specification generation, and the documentation now includes both a reusable template and a completed SONiC linkmgrd example.
• --keep-original now follows project ignore rules while retaining tracked paths and trace metadata. Its Git patch remains reviewable by leaving binary and trace payloads in the private source copy instead of embedding them in changes.patch.
• The repository now has a concise, curated changelog, and the case-studies submodule includes the latest reviewed SONiC, SREGym, Apache Ratis, libspdm, and CometBFT artifacts.

Upgrade notes

After updating an existing checkout, rerun specula setup to refresh the installed skills and tools. Codex plugin users should select plugin again to update specula-codex@specula.

If you use the curated case-study corpus or submodule-hosted research tools, also run git submodule update --init --recursive.

Installation limitations

v1.1.0 remains a source release. Plain wheel and PyPI installation are not supported because the runtime depends on repository-local scripts, skills, and tools. Install from a persistent checkout with uv tool install -e . and keep that checkout in place.

GitHub-generated source archives do not contain submodule contents. Use a recursive Git clone when the curated case-study corpus or submodule-hosted research tools are needed.

Known limitations

• Native Windows is not supported; use WSL2.
• Exact-session resume requires provider-persisted conversation state. It cannot restore a terminated CLI process's in-memory state or an in-flight child process; use --fresh-context when the saved conversation cannot be resumed.
• --keep-original requires isolated output mode and rejects unsafe layouts such as external symlinks, linked worktrees, submodules, nested repositories, or externally shared Git objects. Without the optional sandbox, it cannot prevent deliberate writes to the original checkout through an absolute path.
• Some frontier models require additional provider access for Phase 4 bug confirmation; use hybrid agent configuration when a different model or provider is needed for that phase.

Full changelog: v1.0.0...v1.1.0