Repository navigation
Releases
v0.3.1
Added
-
Colored agent roster shared by shell and TUI: required CORE roles, enabled ON
and disabled OFF workers, short roles and LLM/tool-only execution. Keep every
worker visible in a compact pinned grid;/agents NAMEopens manifest details
and toggle hints, while/agents architectureretains the full runtime view. -
Local Coder assignments use ProtoLink 0.8.2 checkpointed work slices and
structuredRunHandoffprogress. The application grants bounded continuation
only for fresh applied edits with current revisions; the SDK preserves the
child task, history, corrections, committed actions and cumulative budgets. -
Existing approval and question controls supply exact responses through native
resume_task, retaining streaming after approval, input and work handoffs.
Execution checkpoints are private temporary files for the active workflow;
the coding Graph and TaskRecord are not reconstructed after process exit. -
PROTOAGENT_CHECKPOINTED_WORKERS=0selects live fallback limits locally.
Explicit transport meshes retain that fallback. The existing
PROTOAGENT_CODER_MAX_WINDOWSbounds progressing slices per assignment.
Changed
- Simplify verification to one harness-controlled check pass after editing, with
no default baseline, model-managed test sequencing or automatic repair loop.
Explicit test requests use no model inference. Missing setup preserves edits
as unverified with a compact note; no install conversation or retries follow. - Bind the entire workflow, checkpoint resumes and final checks to one native
shared budget. Load only embedded-Python-compatible virtualenv packages. - Add
/agents verifier on|offand its shell equivalent. Off skips check
discovery and commands. Tester is opt-in for new configurations; saved user
settings are preserved. Useful tests are encouraged without requiring generated
test files solely to satisfy a harness gate. - Require ProtoLink 0.8.2 and register narrow worker interfaces with the SDK's
Agent.add_worker_tool. Remove custom delegation wrappers and command preflight
from the controller; original workers retain their own policies and approvals. - Consume immediate child events from the public native stream, removing the
separate Task-metadata polling loop. - Raise source and command preparation rejections directly into ProtoLink's
correction loop before approval/execution. Remove executable validation
actions; rejected preparation consumes no tool call. - Remove the application Anthropic SDK upper bound now that the engine supports
SDK v1. Keep provider compatibility checks on real SDK serialization with
local HTTP fixtures. Forward configured Gemini, OpenAI and Anthropic API URLs. - Update CLI, core and documentation metadata together, plus the Guide,
installation/development instructions, engine integration docs and whitepaper.
v0.3.0
HUGE Update
Dependency maintenance
Refresh Rust dependencies, including PyO3 0.29.3, Crossterm 0.29, Inquire 0.9.4, Indicatif 0.18.6, Console 0.16.6 and Tokio 1.53.3. Update the Cargo lockfile and validate with stable Rust 1.99.
Pin Ruff 0.17.0, ty 0.0.87 and build 1.6.1 in requirements-dev.txt for consistent contributor and CI checks. Update installed Python runtime integrations, retaining ProtoLink 0.8.1 as the engine. Bound Anthropic below 1.0 because its new Messages API removes generation parameters still sent by ProtoLink's adapter.
Update Docusaurus to 3.10.2, React to 19.3.1 and the docs lockfile. Retain ELK 0.1.9 as required by the Docusaurus Mermaid theme.
Refresh GitHub Actions and use Node 24 for documentation CI.
Added
Compact styled Edits cards in CLI/TUI, with green additions, red removals, bold literal file names and muted diff hints. Verification summaries fit one line, with full logs retained in /trace.
Missing Python runners/external dependencies end final verification quietly before another model request, preserving applied edits as unverified without install questions or code-repair loops. Genuine regressions retain bounded repair; native failed receipts and stop conditions remain authoritative.
Small-profile Gemma 4 automatic action acquisition uses ProtoLink's JSON protocol when advertised native tooling emits narration/channel/action text. Explicit native/json settings remain authoritative. Missing verification receipts are labeled not run, stopped-before-edit results report no changes, and Python dependency/runner failures get environment/setup explanations.
Parse-only Python afterimage guards for create, replace and exact edits reject introduced syntax errors before approval through native no-change feedback. Literal escapes remain unchanged; already malformed sources can be repaired incrementally, and valid proposals retain native write authorization/recovery.
Per-project /approvals picker and proto-cli approvals controls: Ask (default), automatic file edits, or automatic edits and commands. Exact native broker decisions retain policies, revision checks, checkpoints and audit records; MCP access and restoration still ask. The TUI displays the active mode.
Deterministic end-of-run applied-file summaries and combined diffs from native snapshots, including partial failures. /diff shows applied changes and /diff proposed labels proposal previews separately. Pre-existing work, rejected proposals, external edits and uncertain effects are distinguished.
Model-free project inspection in shell/TUI: ls, browse/files, open and cat. Folder browsing supports filtering, parent navigation and Enter to view text; the read-only viewer has line numbers, paging and horizontal panning. Listings, source loading and transcript previews are bounded; paths remain in the active project, and terminal control characters are displayed inertly. /open now views files; /project selects a workspace. Guide and completion share the catalog.
Runtime-owned TaskRecord with criteria, check selection, scoped worker packets and explicit done/needs_context/blocked worker reports.
Stateless, read-only Tester for regression planning and failure analysis.
Persistent optional-agent controls for Tester (default on), Scout and MCP (default off). Disabled workers are not constructed or registered; required Architect/Explorer/Coder/Verifier roles cannot be disabled.
mcp and /mcp setup, explicit discovery probes, server/tool allowlists and environment-referenced HTTP authentication. Model-free MCP broker with three fixed discovery/schema/call tools, native approvals and bounded model handoffs.
Repository check discovery, .protoagent/project.json configuration, and Verifier run_check(check_id, phase) with baseline and final verification.
Coder bounded source reads and exact single-occurrence edit_file with a source revision guard, native diff approval and native recovery.
Compact small-profile protocol prompts and worker cards with exact input schemas, omitting repetitive examples and large output schemas.
Per-request context admission through native ProtoLink policies and hooks, reserving output tokens and retaining the current task and runtime record. Small profiles with unknown capacity receive an 8192-token application cap.
Disposable coding evals with independent acceptance scripts and a same-model single-agent baseline: proto-cli eval coding --plan|--live.
Owned local subagents through ProtoLink 0.8: independent conversations, inherited policies, shared Graph budgets, sequential dispatch, depth one and configurable per-attempt child limits. Default execution needs no Registry server or loopback sockets; explicit transport meshes remain available.
Native ContextPolicy and AgentHooks for current task state, complete-turn pruning and scoped progressive retrieval of large tool/child observations.
Optional explicit same-provider fallback_models through native RoutedLLM, with bounded transient request fallback, shared accounting and no tool replay.
eval harness [--json]: native evaluate() runs repeated offline application read/task-state cases with fresh factories and linked child receipt checks.
Architect-only native ask_user for clarification and same-task continuation. TUI/shell question input supports explicit suggestions, free text, skip and cancellation, correlated replies, native deadlines and no implicit approval.
Changed
Admit up to three progressing native Coder work windows in one assignment, rather than interrupting every nine calls. Each continuation requires fresh applied writes, current revisions and no native stop or explicit missing-context report. Scope, history, receipts, approvals and aggregate budgets are preserved. A stalled window cannot borrow earlier progress. PROTOAGENT_CODER_MAX_WINDOWS lowers the application ceiling (1–3, default three); worker.continuing updates the transient TUI activity. Native inference limits remain unchanged.
Return exhausted partial Coder assignments directly to the bounded Graph before another Architect model call, avoiding routine continuation questions that spend the run deadline. workflow.work_yielded records the transition; fresh Graph nodes can still inspect and narrow remaining work.
Explain native runtime expiry, including expiry while awaiting a user answer, rather than presenting it only as an unresolved worker/check failure.
Return completed failing final checks directly to the native repair Graph before another Architect model request. This prevents edits in a closed check phase from turning an eligible repair into a policy denial. After two repairs, remaining failures finish incomplete. Missing dependency failures can still prompt for a manual verification handoff in the repair attempt.
Deduplicate live native events across the whole workflow, so continuation and enclosing handles do not replay old activity, answer chunks or command output. Complete native reports retain their execution evidence.
Continue stable, applied partial Coder work when Architect returns too early: the native Graph permits at most two continuations before final verification, preserving the frozen plan, edit phase and shared budget. Execution failures, denials, uncertainty, no-progress handoffs and manual verification stops do not qualify. Exhausted continuations report incomplete work without retaining unfinished tool-call prose as the final answer.
Small-profile Ollama requests default to at most 4096 generated tokens; small Gemma 4 variants default to think=false for short coding actions. Explicit num_predict/max_tokens and thinking settings win; sampling is preserved. Generation-repeat aborts now explain the provider failure separately from agent turn limits; partial streamed tool proposals are never executed/replayed.
Statically recognize pytest functions/imports/test classes when selecting a Python runner, including test/ directories. Preserve unittest for its own suites and owner-declared checks. Root pytest checks retain current moved test files/configuration as dependencies; discovery never imports project code or installs a missing test runner.
Enforce Coder's work-slice handoff before the last native inference turn, without requiring a final model reply. A public Agent extension consumes only the harness's pre-model stop signal, preserving native execution and stop conditions. Return recorded partial edits, or a blocked no-progress report. Architect can assign only unfinished work within the shared run budget; unresolved handoffs never satisfy completion. Reject identical exact edits before approval and return the approved source digest after successful edits to avoid redundant reads. Failed runs show applied paths and readable causes while retaining the complete native error and receipts in the trace.
Hide fallback tool/delegation JSON from native-capable model streams, reusing ProtoLink normalization for display classification. Preserve JSON answer data and typed final content; clear provisional narration at native tool selection so the muted work status returns instead of retaining an action as an answer.
Show one muted, transient activity line while an answer is loading. Native work phases replace it in place, distinguish command preparation from running, and disappear when text streams or the run ends. Saved answers retain no loading labels; phase updates survive the live trace preview cap.
Resolve bare command programs and project-relative working directories before native approval, preserving explicit environments and interpreter identity. Default output limits fit the configured ceiling. Small local calls return native no-change correction feedback for invalid preparation before dispatch.
Preserve root unittest discovery when tests move into packages; capture moved tests and package markers as current check dependencies without changing the frozen runner. Agent...
v0.2.3
Added
Live model generation and delegated command stdout/stderr in shell and TUI, using native RunHandle.events() and explicit streaming Agent capabilities. Answers stream under a consistent AGENT / architect or AGENT / guide heading with a cyan prompt and steady mint _ cursor. Runtime activity stays in the status area; /trace exposes details and up to four retained worker/process previews of 4096 characters. Answers stay complete.
Streaming Guide help in /help QUESTION and proto-cli help "QUESTION", with native cancellation and no project or saved conversation required.
Packaged command reference shared by Guide and TUI command completion, including /config, shell equivalents, settings commands and aliases.
/debug [on|off] reveals or hides response metadata and a /trace hint, including on existing answers. It defaults off for each TUI session.
Animated three-dot thinking, faint hints inside the empty composer, and a slow cursor blink (700 ms per phase) that does not repaint the chat.
Two muted composer borders and a lower single-line prompt that expands upward for multiline input; command suggestions sit in the lower border.
Markdown styling during generation for Architect and Guide answers: bold, italic, headings, highlighted inline code and indented fenced code blocks. Formatting survives wrapped lines and partial delimiters; stored answers retain their original Markdown.
Shell Ctrl-C requests native cancellation and waits for runtime cleanup.
Provider-free gated streaming tests for early HTTP model delivery, JSON-action and native-tool modes, cancellation, cleanup, live delegated process output, receipt deduplication, redacted persistence and checkpoint pagination.
Changed
Require ProtoLink 0.7.1; coordinate CLI/core/docs versions at 0.2.3.
Consume propagated worker receipts directly from native parent/Graph tasks. Remove ApplicationRunStore.trace_report() and stored-worker evidence scans.
Configure SQLiteRunStore(..., redaction_policy=...) for automatic task, report and caller metadata writes. Remove save overrides and custom credential replacement in favor of native RedactionPolicy.sensitive_values.
Use native checkpoint list_changes() with state/run filters and pagination. Remove the application inventory reader over the raw Storage namespace.
Read progress JSONL incrementally by byte offset, retaining partial writes for the next poll. Live output is separate from the bounded trace-summary channel.
Cache immutable message layouts and repaint only changed transcript rows during streaming. Animation ticks retain the answer without copying or reformatting it; resize, debug and modal transitions refresh the relevant layout.
Bound live TUI ingestion to 256 records or 256 KiB checked between records per poll, retain the latest 512 status summaries, and drain remaining records at completion. Native traces and complete answers remain intact.
Update docstrings, runtime readiness checks, Guide help and documentation.
Fixed
Stop dropping model chunks before they reach the Rust frontend. An HTTP worker mesh no longer disables live output from the locally invoked Architect.
Keep worker/step previews separate, replace final text without duplication and retain valid UTF-8 when bounding previews or reading partial progress lines.
Hide JSON-action envelopes during generation: progressively decode final answer text, including split escapes and Unicode, into the Architect message. Native-tool models can still stream genuine JSON answers unchanged. Trace details remain available through /trace without repeating the answer.
Mask configured secrets split across live output chunks and strip terminal controls from presentation. Unknown secrets still require application handling.
Preserve failed, uncertain and input-required labels in the TUI.
Stop the answer jumping at completion: retain the same message layout and cursor cell, avoid inserting report footnotes or collapsing a trace above it, and synchronize terminal redraws where supported.
Keep runtime activity only in the bottom status bar. Remove the duplicate top loading indicator and the bright input instructions beside the answer.
Give the response cursor the same background as its text and stop its blink. Keep the terminal's default background for answers and the existing UI palette; limit Markdown color changes to response text and code highlights. Position styled spans by Unicode display width so emoji and wide characters do not overlap the next span.
Give Guide explicit configuration guidance: /config and proto-cli config inspect redacted settings; model, key, context and agent commands change them.
Boundaries
Streamed answers are provisional until the native task establishes completion, failure, cancellation or uncertainty. Partial JSON is projected for display only; actions remain assembled, authorized and executed by ProtoLink.
PROTOAGENT_STREAM=0 hides previews without changing execution or replaying work.
Existing approvals, budgets, POSIX recovery constraints, legacy inspection and bounded repairs remain. Native inventory includes original recovery bytes; ProtoAgent exposes metadata only and keeps recovery storage private.
v0.2.2
[0.2.2] - 2026-09-12
Changed
- Require ProtoLink 0.7.0 across the core, CI and installation guidance;
align active CLI/core/docs versions to 0.2.2. - Register native
process_tool()on Verifier asexecute_command. Remove the
custom subprocess runner, output drains, timeout and process cleanup code.
Commands use explicit argv, cwd, env and limits; the environment is no longer
inherited. Keep 600-second and 32-KiB application ceilings and native budgets. - Replace manual embedded startup/cleanup and final-event reconstruction with
AgentGroup,RunHandle, normalizedRunResultand native reports. - Use
ApprovalBrokeras the actual handler. Rust previews native JSON command
specifications and filesystem diffs, echoes the exact request ID/fingerprint,
and uses private temporary controls. Application authorization supplies scope. - Register native create/replace/preview/restore filesystem tools with a dedicated
StorageCheckpointStore, private SQLite storage and one writer per project.
Remove custom diff, hash, file mutation and checkpoint execution helpers. - Define completion with native
CompletionCheck/CompletionValidatorover
executed outcomes and current resource revisions. Enforce an initial attempt
and at most two repairs with native Graph limits, separate from transport retries. - Update docstrings, Guide help, the operator manual and migration/API mapping.
Fixed
- Approval, previews and delegation no longer count as a completed write.
- Quality evaluation counts executed native file edits; command approvals are
not mistaken for Coder usage. Runtime diagnostics check the required 0.7 APIs. - Preserve failed, canceled and uncertain results through the frontend adapter;
errors after submission do not imply effects were absent. - Reject stale preimages, restoration conflicts and further mutations after an
uncertain checkpoint. Never automatically replay interrupted effects. - Redact known credential values and recovery bytes from presentation/report
copies while retaining protected native recovery and approval storage.
Migration boundaries
- Recoverable Coder writes now require POSIX, absolute project paths without
symlinks and existing parents. New files use native mode 0600. Directory
creation requires a separately approved command and a later edit run. - Preserve v0.2.1 checkpoint databases as inspection-only legacy records;
they cannot establish native revision identities. Chained undo can conflict
after a later restoration changes a file's identity. - Local commands run on the host without sandbox isolation; recovery covers
Coder file effects only. RunReplay remains inspection, not resumption. - ProtoLink 0.7.0 delegation does not merge worker receipts into parent reports.
Compose same-trace native RunStore snapshots for acceptance. Native snapshot
redaction and checkpoint inventory still need small application adapters.