v0.20.1

A patch release from a whole-codebase bug hunt. Candidate issues were
confirmed by reading or running the code, independently challenged, and fixed
with a regression test each.

Fixed

• Tool-call and tool-result lines no longer parse tool arguments or output as
  Rich markup. Text such as ls [/tmp] used to crash the turn and leave an
  unanswered tool call in history, and code previews silently dropped
  bracketed text (d[key] showed as d). Recalled-context and info lines
  get the same treatment.
• Tool failures reported as Error reading/writing/listing ... are now
  classified as failures instead of "worked".
• search_files keeps ripgrep matches when ripgrep also reports errors (for
  example one unreadable folder) and marks the result as partial. Long
  stderr no longer stops the search early. The Python fallback now honors
  path globs such as **/*.py, dir/*.py and !*.md, case-insensitively on
  Windows.
• run_shell keeps leading whitespace on the first output line, so formats
  such as git status --short stay intact.
• A schemeless OLLAMA_HOST (for example 127.0.0.1:11434) no longer fails
  the readiness check and blocks every prompt, and the supplemental gateway
  receives a normalized host.
• Agent-block models are routed to Ollama Cloud or local Ollama by the block's
  own model instead of the active model.
• --oneshot --json always ends with a done event, reporting partial when
  saving the configuration fails.
• /agent fixes: a model error marks the block as failed; tasks that begin
  with show, switch, resume or fork are only treated as thread
  commands when followed by a thread reference; apostrophes and quotes in the
  task no longer drop --pipeline or --roles; Ctrl+C on a team run stops
  waiting for running specialists.
• Saving a memory never overwrites a memory.json that exists but cannot be
  read safely; an empty file still counts as no memories. One invalid legacy
  fact no longer breaks /remember or turns off recall, and the load error
  clears after a successful repair.
• Lessons written as paragraphs (the template format) are indexed, not only
  ## sections.
• Automatic memory capture rejects complaints and reported speech such as
  "You always forget to run the tests".
• When the context summarizer fails, the lossy fallback summary is reported
  instead of being used silently.
• Code-RAG re-embeds chunks after an embedding-width or embedder change
  instead of failing on every turn, keeps older indexes usable after upgrade,
  and drops deleted files from the repository map.
• A source file with a future modification time no longer forces a full
  harness re-index on every load.
• Ctrl+C during a streamed answer keeps the partial answer in the
  conversation and says so.
• /agent resume and /agent fork no longer treat English words made of hex
  letters (add, beef, decade) as thread references; a reference followed
  by a task must contain a digit or match a full thread id.
• A schemeless OLLAMA_HOST now also works for embedding identity and the
  gateway upstream, so code, lesson and memory retrieval stay semantic.
• File contents that happen to begin like a tool error are no longer recorded
  as failed tool calls, in the chat loop, /agent, one-shot and program runs.
• Partial search results are recorded as not fully successful.
• Malformed search globs such as *[^] no longer crash the Python fallback.
• /agent team cancellation stops write_file before it writes and
  terminates a running run_shell command.
• Ctrl+C during a parallel tool batch returns within a bounded wait and keeps
  every tool call paired with a result.
• /diff, /changes, agent-block panels, /help, the memory table and the
  runtime overview show bracketed text literally instead of treating it as
  Rich markup.
• Existing lesson indexes are rebuilt with the new chunker, and lesson text
  containing comment markers is kept intact.
• Memory capture: needed/embedded standing rules are accepted, and
  first-person complaints (I always forget ...) are rejected.
• Files created after the last code-RAG scan are found on the next scan.
• The settled thinking panel shows both the opening and the ending of long
  reasoning, and the live panel keeps indentation.

Changed

• Ctrl+C at the prompt clears a half-typed line; on an empty line it asks for
  a second Ctrl+C (or Ctrl+D) before exiting.
• The footer reflects /safe, /auto, /theme and /cd changes
  immediately.
• The model-wait spinner shows the elapsed time, and switches to "loading" for
  slow local model loads.
• The live thinking panel follows the newest reasoning instead of freezing
  after the first 1,200 characters.
• Errors show their class and, for connection, timeout and missing-model
  failures, a short hint about what to check.

Security

• The public-release scan also rejects local keychain item names used for
  companion credentials, in source and in built artifacts.

Release Limits

• Native signing, M8 external-browser metrics and other blocked hardening
  work remain blocked and are not claimed.

v0.20.0

Added

• Add explicit, session-only /mode yolo activation with an owner, process,
  and workspace binding. YOLO is never saved, inherited by child agents, or
  enabled by a prompt, copied configuration, or harness refresh.
• Route memory through native Continuum when it is explicitly selected, with
  fail-closed scoped context, revision-bound writes, and no fallback backend.
  algo-cli config memory repair removes retired memory selectors from an
  existing configuration while preserving every other setting.
• Add the optional Jev kernel: jev_question_contract sends bounded,
  advisory question contracts to the separately installed jev-workflows
  companion. Lint runs locally; inference requires algo-cli config jev enable
  and session approval, and validated answers never authorize actions.
• Keep the normal model, context, tool, reflection, and safety status visible
  during generation and tool execution on supported interactive terminals.

Changed

• Let user-activated YOLO preapprove registered in-scope actions while retaining
  safe-mode denials, credential and direct-send restrictions, protected memory
  boundaries, target checks, child isolation, and post-mutation verification.
• Allow curated owner-bound read-only observation outside the active workspace
  in YOLO without extending cross-workspace mutation authority.
• Normalize an empty typed-program output list to the final step and ignore only
  redundant active-workspace cwd fields owned by the runtime. Conflicting
  workspaces and invalid plans remain rejected.
• Remove the ordinary tool-round and verification-recovery ceilings while YOLO
  is active; saved /toolmax behavior is unchanged after leaving YOLO.
• Preserve live YOLO activation through /reload without persisting it, and use
  native cloud-model context limits instead of stale catalog stamps.
• Show a short / menu of common commands, group /help by category, and
  reveal subcommands after their parent command. Existing command names and
  aliases still work.
• Render the in-progress footer with the same content, colors, and spacing as
  the prompt footer, so submitting a question no longer restyles it.

Removed

• Ship docs/ALGO.md as an empty template instead of a populated pattern
  catalog. Each user keeps their own catalog at ~/.algo_cli/ALGO.md, which
  Algo indexes in place of the template; /intelligence init creates it.
• Remove the Acrobat, finance, and construction kernels and their modules from
  the package. Personal kernels are declared in
  ~/.algo_cli/kernels/kernels.json and load on that machine only; /kernel
  lists them separately from built-in kernels.

Fixed

• Stop a startup crash when Continuum's required shared context exceeds the
  default packet size: Algo retries once with the backend-reported size, up to
  a hard cap, without dropping required records.
• Check protected-memory Git paths with their exact bytes, so a hard link with a
  non-UTF-8 name cannot bypass the link check, and never launch a
  continuum-memory command found inside the current repository.
• Stream Continuum replies under a fixed size limit, keep /memory help
  available when the backend is down, and bound the personal kernel file.
• Keep a context window set with /ctx even when it equals a common model
  default; only unrecorded automatic values are promoted to the native window.
• Refresh the Boron managed-browser pin to Chrome 153.0.8010.52 with its
  Debian package checksum, release timestamp, regenerated dpkg lock, and
  coupled fixtures, so hosted browser qualification is within the 72-hour
  security-update limit.
• Distinguish compiler-format failures from missing runtime authority so valid
  YOLO programs are not reported as approval failures.
• Classify slash-command inspection by each handler's real argument contract.
  Mutating /host, /model, /system, /goal, /google, and related forms
  no longer inherit read-only output capture or inspection treatment.
• Preserve /memories output capture and keep mutating command output on the
  normal non-captured policy path.
• Restore terminal scroll margins on normal completion, resize, suspension,
  termination, and hangup; suppress terminal controls for JSON output and
  terminals without VT support.
• Accept both current owned Swift test-bundle identities without weakening path
  or process-ancestry checks, and update the dependency lock past known anyio
  advisories.
• Pin release upgrade qualification to the actual published 0.19.2
  predecessor and its exact public wheel bytes instead of the obsolete
  0.18.0 baseline, including the real published updater for pipless
  uv-pip installations instead of a manual bootstrap bypass.

Release Limits

• Retired memory backends (Echo Veil and D-57) are removed, not supported as
  options or fallbacks. Their historical records are left unchanged.
• M8 retains five blocked external-browser metrics and M9 retains thirteen
  blocked requirements. This release does not claim external-browser or model-
  quality qualification from local deterministic evidence.

Algo CLI 0.19.2

Qualified reliability and release-delivery update.

• Repairs standalone uv-pip self-updates without requiring a pip module, while preserving uv-tool ownership.
• Adds bounded PyPI visibility retries and rejects malformed or incomplete registry metadata.
• Refreshes the managed-browser security pin and hosted qualification inputs for Chrome 153.0.8010.36.
• Separates immutable tagged source from the current protected-main publisher during exact-asset recovery.
• Includes the reliability, security, model-discovery, Windows, and state-preservation fixes accumulated after 0.19.1.post1.

Historical v0.19.0 and v0.19.1 drafts remain unpublished. This draft is published only by the protected release workflow after source, artifact, approval, attestation, and PyPI checks pass.

Algo CLI 0.19.1.post1

Packaging-compatible post-release for the qualified 0.19.1 candidate.

• Removes the unpublished direct Git dependency from public package metadata while retaining the exact private qualification pin.
• Includes the frozen reliability, security, model-discovery, and Windows fixes.
• Includes exact browser-broker accounting, verifier precision, and scheduler-safe qualification corrections.

This draft is published only by the protected release workflow after source, artifact, approval, attestation, and PyPI checks pass.

Algo CLI v0.18.0

Algo CLI v0.18.0

This release makes harness readiness harder to claim and easier to inspect.

Highlights

• Indexes 84 ActionSpec tools, commands, providers, and kernels as searchable runtime capabilities with their policy, prerequisites, compatibility, and limitations.
• Adds a frozen multilingual and complex-query retrieval gate covering code-switching, translated aliases, multi-hop retrieval, cross-source reconciliation, and temporal supersession.
• Requires recall@k, MRR, nDCG, citation precision, and a fixture digest for the harness retrieval scorecard to pass.
• Reports privacy-safe external-store availability, provenance, conflict handling, malformed configuration, and permission diagnostics.
• Recovers invalidated ChatGPT/Codex OAuth sessions with one bounded refresh and actionable sign-in guidance.
• Hardens loopback readiness caching, web-fetch concurrency ownership, and terminal configuration suggestions.
• Adds operational guides for external stores, provider authentication recovery, runtime capabilities, and Echo Veil readiness.

Measured release evidence

• 1,939 tests passed locally.
• All GitHub CI, native-helper, website, distribution, and installed-wheel jobs passed on Linux, macOS, and Windows.
• Live local scorecard: 10/10 with 241/241 embeddings and 84 runtime-capability records.
• Frozen retrieval workload: recall@4 1.0, MRR 1.0, nDCG@4 1.0, citation precision 1.0.
• Public source and distribution scans passed.

Honest boundaries

• External agent stores remain disabled by default and require explicit opt-in.
• Echo Veil is reported unavailable unless its package and complete runtime path are genuinely wired.
• The frozen quality workload measures offline retrieval, not universal model superiority or production-scale answer generation.

Upgrade with:

algo-cli update

Or install directly:

pipx install algo-cli-runtime==0.18.0

Algo CLI v0.17.0

Highlights

• Adds terminal-first algo-cli config setup for ChatGPT/Codex, Google Workspace, and xAI.
• Migrates xAI runtime access to the documented API-key route.
• Hardens OAuth refresh persistence, credential endpoints, shell timeout cleanup, bounded web fetches, runtime state loading, and X-search caching.
• Ships synchronized CLI, documentation, tests, and website metadata for v0.17.0.

Validation

All 10 protected CI checks passed across Linux, macOS, and Windows, including source/privacy scans, 1,912 tests, package inspection, and installed-wheel smoke tests.

See CHANGELOG.md for the complete release notes.

Algo CLI v0.16.0

What's Changed

• Release v0.16.0 by @Seabass-up in #17

Full Changelog: v0.15.1...v0.16.0

Algo CLI v0.15.1

What's Changed

• Add self-update command by @Seabass-up in #15
• Release v0.15.1 by @Seabass-up in #16

Full Changelog: v0.15.0...v0.15.1

Algo CLI v0.14.0

Algo CLI v0.14.0 is the first sanitized public release of the agent runtime.

Highlights

• persistent runtime-agent threads, bounded multi-agent teams, and kernel wiring checks
• privacy-gated automatic memory capture and durable context controls
• execution guardrails, Git evidence, performance telemetry, and verification contracts
• packaged algorithm, memory, wiki, and skill corpus
• public-release privacy scans, cross-platform CI, and isolated-wheel smoke tests

Install

The PyPI distribution is named algo-cli-runtime; it installs the algo-cli command.

pipx install algo-cli-runtime
algo-cli doctor

The release workflow publishes with short-lived GitHub OIDC credentials through the protected pypi environment. No long-lived PyPI token is used.

See CHANGELOG.md for the complete release notes.