misterdev 0.3.1__tar.gz → 0.4.1__tar.gz

{misterdev-0.3.1/misterdev.egg-info → misterdev-0.4.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: misterdev
-Version: 0.3.1
+Version: 0.4.1
 Summary: An autonomous, extensible LLM build orchestrator that plans, edits, and verifies code across languages.
 Author-email: David Condrey <david@writerslogic.com>
 License-Expression: AGPL-3.0-or-later
@@ -51,6 +51,7 @@ Provides-Extra: mcp
 Requires-Dist: mcp>=1.0.0; extra == "mcp"
 Dynamic: license-file
 
+<!-- mcp-name: io.github.dcondrey/misterdev -->
 <p align="center">
   <img src="https://raw.githubusercontent.com/dcondrey/misterdev/main/assets/logo.gif" alt="misterdev" width="220">
 </p>
@@ -150,7 +151,7 @@ misterdev keeps a durable, fingerprinted stream of its own real failures and run
 
 On the correctness side, misterdev works **reproduction-first**: for an issue-driven task it synthesizes a failing test from the acceptance criteria, **validates that the test actually fails on the clean tree** (a test that reproduces nothing is discarded rather than trusted), then drives the fix to turn it green — so the model edits toward a concrete, verified target instead of a description.
 
-**Two-timescale evolution** *(design; see [docs/two-timescale-evolution.md](docs/two-timescale-evolution.md))* takes the self-improvement further than a memoryless runtime agent can: the agent invents task-specific tools at runtime, and the ones that **prove they generalize** are consolidated — through the same held-out promotion gate — into a **persistent, best-per-capability tool library that future runs start from**. Fast loop explores; slow loop keeps the winners; the held-out gate keeps the library general rather than benchmark-overfit. Capability compounds across runs instead of being reinvented each task.
+**Two-timescale evolution** *(built; opt-in via `orchestrator.runtime_tooling`; see [docs/two-timescale-evolution.md](docs/two-timescale-evolution.md))* takes the self-improvement further than a memoryless runtime agent can. At **runtime**, the model may author a small task-specific helper tool that runs **sandboxed** (a hardened, network-less container — untrusted code never touches the host or the repo, and with no container engine the capability degrades off); its output feeds the edit. Every invented tool is then captured with the task's outcome into a **tool corpus** — a free byproduct of normal runs — and a deliberate promotion pass admits the tools whose success **generalizes** on a held-out task split (baseline drawn from the reproduction corpus) into a **persistent, best-per-capability tool library that future runs start from**. Fast loop invents; slow loop keeps only the winners; the same held-out gate that guards scaffold self-edits keeps the library general rather than benchmark-overfit. Capability **compounds across runs** instead of being reinvented each task — the memory the current top open-source scaffolds lack. Run the promotion pass with `python -m misterdev.core.evolution.tool_promotion <project>`.
 
 ### Extensibility
 
@@ -170,7 +171,7 @@ Gate-verified pass@1 on [Aider's polyglot benchmark](https://github.com/Aider-AI
 | Python | 8 / 10 | **80%** |
 | Rust | 7 / 10 | **70%** |
 
-A continuous stress run has solved **20/20** across the three languages with zero failures — including the exercises usually cited as hard (bowling, forth, arbitrary-precision decimal). Every solve is judged by the exercise's own hidden tests, not the model's say-so. Full numbers, methodology, and how to reproduce: **[docs/benchmark-results.md](docs/benchmark-results.md)**. Test suite: **1,897 passing** — **[docs/TESTING.md](docs/TESTING.md)**.
+A continuous stress run has solved **20/20** across the three languages with zero failures — including the exercises usually cited as hard (bowling, forth, arbitrary-precision decimal). Every solve is judged by the exercise's own hidden tests, not the model's say-so. Full numbers, methodology, and how to reproduce: **[docs/benchmark-results.md](docs/benchmark-results.md)**. Test suite: **1,941 passing** — **[docs/TESTING.md](docs/TESTING.md)**.
 
 ## CLI reference
 

{misterdev-0.3.1 → misterdev-0.4.1}/README.md

@@ -1,3 +1,4 @@
+<!-- mcp-name: io.github.dcondrey/misterdev -->
 <p align="center">
   <img src="https://raw.githubusercontent.com/dcondrey/misterdev/main/assets/logo.gif" alt="misterdev" width="220">
 </p>
@@ -97,7 +98,7 @@ misterdev keeps a durable, fingerprinted stream of its own real failures and run
 
 On the correctness side, misterdev works **reproduction-first**: for an issue-driven task it synthesizes a failing test from the acceptance criteria, **validates that the test actually fails on the clean tree** (a test that reproduces nothing is discarded rather than trusted), then drives the fix to turn it green — so the model edits toward a concrete, verified target instead of a description.
 
-**Two-timescale evolution** *(design; see [docs/two-timescale-evolution.md](docs/two-timescale-evolution.md))* takes the self-improvement further than a memoryless runtime agent can: the agent invents task-specific tools at runtime, and the ones that **prove they generalize** are consolidated — through the same held-out promotion gate — into a **persistent, best-per-capability tool library that future runs start from**. Fast loop explores; slow loop keeps the winners; the held-out gate keeps the library general rather than benchmark-overfit. Capability compounds across runs instead of being reinvented each task.
+**Two-timescale evolution** *(built; opt-in via `orchestrator.runtime_tooling`; see [docs/two-timescale-evolution.md](docs/two-timescale-evolution.md))* takes the self-improvement further than a memoryless runtime agent can. At **runtime**, the model may author a small task-specific helper tool that runs **sandboxed** (a hardened, network-less container — untrusted code never touches the host or the repo, and with no container engine the capability degrades off); its output feeds the edit. Every invented tool is then captured with the task's outcome into a **tool corpus** — a free byproduct of normal runs — and a deliberate promotion pass admits the tools whose success **generalizes** on a held-out task split (baseline drawn from the reproduction corpus) into a **persistent, best-per-capability tool library that future runs start from**. Fast loop invents; slow loop keeps only the winners; the same held-out gate that guards scaffold self-edits keeps the library general rather than benchmark-overfit. Capability **compounds across runs** instead of being reinvented each task — the memory the current top open-source scaffolds lack. Run the promotion pass with `python -m misterdev.core.evolution.tool_promotion <project>`.
 
 ### Extensibility
 
@@ -117,7 +118,7 @@ Gate-verified pass@1 on [Aider's polyglot benchmark](https://github.com/Aider-AI
 | Python | 8 / 10 | **80%** |
 | Rust | 7 / 10 | **70%** |
 
-A continuous stress run has solved **20/20** across the three languages with zero failures — including the exercises usually cited as hard (bowling, forth, arbitrary-precision decimal). Every solve is judged by the exercise's own hidden tests, not the model's say-so. Full numbers, methodology, and how to reproduce: **[docs/benchmark-results.md](docs/benchmark-results.md)**. Test suite: **1,897 passing** — **[docs/TESTING.md](docs/TESTING.md)**.
+A continuous stress run has solved **20/20** across the three languages with zero failures — including the exercises usually cited as hard (bowling, forth, arbitrary-precision decimal). Every solve is judged by the exercise's own hidden tests, not the model's say-so. Full numbers, methodology, and how to reproduce: **[docs/benchmark-results.md](docs/benchmark-results.md)**. Test suite: **1,941 passing** — **[docs/TESTING.md](docs/TESTING.md)**.
 
 ## CLI reference
 

misterdev-0.4.1/man/misterdev.1

@@ -0,0 +1,142 @@
+.TH MISTERDEV 1 "2026-07-11" "misterdev" "User Commands"
+.SH NAME
+misterdev \- autonomous LLM build orchestrator
+.SH SYNOPSIS
+.B misterdev
+[\fIcommand\fR] [\fIoptions\fR]
+.br
+.B misterdev
+\(dq\fIplain-english request\fR\(dq
+.br
+.B misterdev
+.SH DESCRIPTION
+.B misterdev
+points a large language model at a repository and a goal, reads the codebase as
+a symbol graph, decomposes the goal into ordered tasks, edits code with anchored
+SEARCH/REPLACE hunks, and runs every change through correctness gates (build,
+lint, tests, typecheck) before reporting done. A change that regresses the test
+suite is reverted through git; nothing merges unless it stays green.
+.PP
+There are three ways to drive it, from most guided to most explicit:
+.TP
+.B Interactive
+Run
+.B misterdev
+with no arguments (or
+.BR "misterdev i" )
+for a guided menu \(em pick an action, answer a couple of prompts, confirm.
+Nothing to memorize.
+.TP
+.B Natural language
+If the first argument is not a known command, the whole line is treated as plain
+English:
+.B misterdev
+maps it to an action with its own model, previews it, and asks before anything
+mutating.
+.TP
+.B Flags
+The subcommands and options below, for scripts and power users.
+.SH COMMANDS
+.TP
+.BR interactive ", " i
+Launch the guided interactive menu (also the no-argument default).
+.TP
+.B build \fR[\fIpath\fR] [\fIgoal\fR] [options]
+Autonomously build, debug, or complete a goal in a project. The goal is free
+text; the words \fBdebug\fR or \fBcomplete\fR select those modes.
+.TP
+.B run \fR[\fIpath\fR] [options]
+Execute a project's pending tasks with dependency-aware ordering, parallel
+waves, and progress-based resume.
+.TP
+.B plan \fR[\fIpath\fR] [options]
+Analyze the project, recommend work, and compose a plan interactively.
+.TP
+.B status \fR[\fIpath\fR]
+Show a project's tasks and their state.
+.TP
+.B report \fR[\fIpath\fR]
+Summarize the latest build's cost, per-model ledger, and audit trail
+(read-only).
+.TP
+.B scan \fIdir\fR
+Discover and register projects (those with a \fBproject.yaml\fR) under a
+directory.
+.TP
+.B list
+List registered projects.
+.SH OPTIONS
+.SS build
+.TP
+.BI \-\-budget " DOLLARS"
+Cap the run's spend (default 100).
+.TP
+.B \-\-dry-run
+Plan only; show what would be done without editing.
+.TP
+.BI \-\-max-tasks " N"
+Cap the number of decomposed tasks.
+.TP
+.B \-\-no-verify
+Skip the final quality-gate chain.
+.SS run
+.TP
+.BI \-\-tasks " FILE"
+Execute an external task-list file instead of the \fBdevplan/\fR directory. The
+file may live in another repository and be in any format \(em JSON, YAML,
+Markdown, or plain text; ordered or unordered lists; multi-line tasks; phases;
+and a dependency table (\fB| Task | Blocked By |\fR) that unlocks parallel
+execution.
+.TP
+.BI \-\-task " ID"
+Run a single task by id.
+.TP
+.B \-\-dry-run
+Show the execution plan (topological waves) without running.
+.TP
+.B \-\-status
+Show which tasks would run vs. skip, then exit.
+.TP
+.B \-\-force
+Re-run all tasks, ignoring the completion cache.
+.SH EXAMPLES
+.TP
+Guided menu (no flags):
+.B misterdev
+.TP
+Plain English:
+.B misterdev \(dqfix the flaky auth test but keep it cheap\(dq
+.TP
+Explicit build with a budget:
+.B misterdev build . \(dqadd rate limiting to the public API\(dq \-\-budget 5
+.TP
+Run a hand-written task list from another repo, previewing first:
+.B misterdev run ./app \-\-tasks ../plans/PLAN.md \-\-dry-run
+.TP
+Fix everything that is broken:
+.B misterdev build . debug
+.SH ENVIRONMENT
+.TP
+.B OPENROUTER_API_KEY
+Default API key variable (overridable via \fBllm.api_key_env_var\fR in
+\fBproject.yaml\fR). The key is never written to config on disk.
+.SH FILES
+.TP
+.B project.yaml
+Per-project configuration (language, build/test/lint commands, gates, MCP,
+targets, \fBtasklist\fR). See \fBdocs/configuration.md\fR.
+.TP
+.B devplan/
+Directory of pre-written task files (Markdown + YAML front-matter).
+.TP
+.B .orchestrator/
+Per-project run state: progress cache, contracts, worktrees, audit trail.
+.TP
+.B ~/.misterdev/registry.json
+Registered-project cache. Stale entries (no \fBproject.yaml\fR) are pruned on
+load.
+.SH SEE ALSO
+Project documentation under \fBdocs/\fR: getting-started, configuration, mcp,
+plugins. Homepage: https://github.com/dcondrey/misterdev
+.SH AUTHOR
+David Condrey.

{misterdev-0.3.1 → misterdev-0.4.1}/misterdev/__init__.py

@@ -1,3 +1,3 @@
 """misterdev — an autonomous, extensible LLM build orchestrator."""
 
-__version__ = "0.3.1"
+__version__ = "0.4.1"