mindweave 2.4.9 → 2.5.1

package/README.md

@@ -1,144 +1,145 @@
-<h1 align="center">Mindweave (mwcode)</h1>
-
-<p align="center">
-  A fast, model-adaptive, terminal-native AI coding agent.<br>
-  You bring your own key. It runs on your machine. Nothing goes anywhere else.
-</p>
-
-<p align="center">
-  <a href="LICENSE">Apache 2.0</a> &nbsp;•&nbsp;
-  <a href="https://x.com/mindweavecli">X</a> &nbsp;•&nbsp;
-  <a href="https://mindweavedev.netlify.app/">Website</a>
-</p>
-
----
-
-## What it is
-
-Mindweave is a coding agent that lives in your terminal and works inside your repository:
-reading, searching, editing, running commands, and checking its own work.
-
-It runs entirely on your machine. There is no backend, no telemetry and no account. Your
-code and your API key never reach a Mindweave server, because there isn't one.
-
-It is built lean on purpose. Most of a coding agent's context budget goes on scaffolding
-the model never needed. Mindweave keeps prompts thin and leaves the room for the model to
-reason about your code.
-
-
-## Mindweave 1
-
-Mindweave 1 is out, tagged `mindweave-1`. It is on npm, and the website is up. Install
-with the steps below.
-
-What landed in it is in the [changelog](CHANGELOG.md): fourteen model providers, a
-rebuilt terminal interface, reworked prompt caching and token accounting, project notes
-the agent maintains across sessions, and a long list of things that were quietly wrong.
-
-## Install
-
-Requires **Windows** and **Node.js 20+**. macOS and Linux are not supported yet, and the
-reason is written down in [KNOWN-ISSUES.md](KNOWN-ISSUES.md) rather than glossed over.
-
-```bash
-npm install -g mindweave
-```
-
-Then, in any project:
-
-```bash
-cd your-project
-mindweave          # or mw, for short
-```
-
-To build from source instead — for development, or to run an unreleased change:
-
-```bash
-git clone https://github.com/mindweave-cli/Mindweave
-cd Mindweave
-npm install
-npm run build
-npm link          # makes the `mindweave` command available globally
-```
-
-It asks for an API key on first launch and saves it to `~/.mindweave/.env`, so every
-project afterwards just works. `mindweave --help` covers the launch flags; everything
-else is configured inside a session.
-
-
-## What it can do
-
-Short version, one line each. The depth is in the linked pages.
-
-- **15 providers, 53 models, one key** — DeepSeek, Anthropic, OpenAI, Gemini, xAI,
-  Mistral, Groq, Cerebras, Qwen, Kimi, GLM, Meta, MiniMax, Tencent, plus every
-  tool-capable model on OpenRouter. Only the driver you
-  use is loaded. Switch with `/provider` and `/model`; remembered per project.
-  [PROVIDERS.md](src/drivers/PROVIDERS.md)
-- **Real tools** — read and edit files, multi-file edits, ripgrep search, a shell with
-  background jobs, and sub-agents. Every edit comes back with the language server's own
-  errors for that file. Read-before-edit is enforced, and `/undo` is a real net.
-- **Deterministic code intelligence** — a background lane indexes your repo with
-  tree-sitter and language servers, costing no tokens, so the agent understands the
-  codebase, not just the open file.
-- **Session memory** — automatic compaction plus a running state summary that survives
-  it, and the agent can read its own earlier sessions in a project.
-- **Project notes** — MINDWEAVE.md is loaded every session and maintained by the agent;
-  `@./path` imports split it, a per-folder one applies only there, and `~/.mindweave/`
-  holds machine-wide notes. `/init` writes the first one.
-- **MCP servers** — connect external tool servers with `/mcp add` or in plain words;
-  their output is untrusted by default. [docs/MCP.md](docs/MCP.md)
-- **Images and web search** — drop a screenshot or write `@shot.png`; ask about a recent
-  release and it looks it up, through your own provider.
-- **Seeing your app** — it can capture one named window (never the whole screen, and it
-  asks first) to tell an app that started from one that works.
-- **Per-project governor** — standing rules, reusable skills, and forbidden paths or
-  commands the agent must respect.
-
-## Using it
-
-| Command | What it does |
-| --- | --- |
-| `/help` | Lists every command |
-| `/init` | Writes MINDWEAVE.md, the project notes loaded every session |
-| `/provider` · `/model` · `/think` | Who answers, which model, how hard it reasons |
-| `/clear` · `/continue` | Start fresh, or resume an earlier session |
-| `/undo` | Reverts what the last turn changed |
-| `/update` | Installs the newest version and reopens on this conversation |
-| `/mcp` | Manages connected MCP servers |
-| `shift-tab` | Cycles interaction modes |
-
-Type while it works and your message queues; press up to take it back and edit it.
-
-## Read more
-
-| | |
-| --- | --- |
-| [CHANGELOG.md](CHANGELOG.md) | What changed, and what each fix actually was |
-| [KNOWN-ISSUES.md](KNOWN-ISSUES.md) | What is broken or unfinished, written down rather than carried quietly |
-| [PHILOSOPHY.md](PHILOSOPHY.md) | How the project is run and what gets into the core. Short, and the honest version |
-| [BOUNDARY.md](BOUNDARY.md) | What belongs in the core versus a driver |
-| [CONTRIBUTING.md](CONTRIBUTING.md) | Mechanics, and what is most useful to report |
-| [SECURITY.md](SECURITY.md) | Reporting a vulnerability |
-| [docs/MCP.md](docs/MCP.md) | Connecting and trusting MCP servers |
-
-## Contributing
-
-Contributions are welcome, especially **model drivers**: Ollama is a stub waiting for
-someone, and a driver is a smaller job than people expect, owning one provider's wire
-format and nothing about how the agent behaves.
-
-Small fixes and reproduced bugs with a failing test can go straight to a pull request.
-For anything larger, start a Discussion first. AI-assisted contributions are fine; the
-one rule is that you understand and have tested what you are submitting.
-
-Bug reports are genuinely the most useful thing you can send. Mindweave is developed by
-running it on real projects and fixing what breaks, and nearly everything in the
-changelog started as a failure someone watched happen. [What to include, and what is
-especially worth reporting.](CONTRIBUTING.md#reporting-a-bug)
-
-Questions are welcome at zallinimann@gmail.com
-
-## License
-
-[Apache License 2.0](LICENSE).
+<h1 align="center">Mindweave (mwcode)</h1>
+
+<p align="center">
+  A fast, model-adaptive, terminal-native AI coding agent.<br>
+  You bring your own key. It runs on your machine. Nothing goes anywhere else.
+</p>
+
+<p align="center">
+  <a href="LICENSE">Apache 2.0</a> &nbsp;•&nbsp;
+  <a href="https://x.com/mindweavecli">X</a> &nbsp;•&nbsp;
+  <a href="https://mindweavedev.netlify.app/">Website</a>
+</p>
+
+---
+
+## What it is
+
+Mindweave is a coding agent that lives in your terminal and works inside your repository:
+reading, searching, editing, running commands, and checking its own work.
+
+It runs entirely on your machine. There is no backend, no telemetry and no account. Your
+code and your API key never reach a Mindweave server, because there isn't one.
+
+It is built lean on purpose. Most of a coding agent's context budget goes on scaffolding
+the model never needed. Mindweave keeps prompts thin and leaves the room for the model to
+reason about your code.
+
+
+## Mindweave 1
+
+Mindweave 1 is out, tagged `mindweave-1`. It is on npm, and the website is up. Install
+with the steps below.
+
+What landed in it is in the [changelog](CHANGELOG.md): fourteen model providers, a
+rebuilt terminal interface, reworked prompt caching and token accounting, project notes
+the agent maintains across sessions, and a long list of things that were quietly wrong.
+
+## Install
+
+Requires **Windows** and **Node.js 20+**. macOS and Linux are not supported yet, and the
+reason is written down in [KNOWN-ISSUES.md](KNOWN-ISSUES.md) rather than glossed over.
+
+```bash
+npm install -g mindweave
+```
+
+Then, in any project:
+
+```bash
+cd your-project
+mindweave          # or mw, for short
+```
+
+To build from source instead — for development, or to run an unreleased change:
+
+```bash
+git clone https://github.com/mindweave-cli/Mindweave
+cd Mindweave
+npm install
+npm run build
+npm link          # makes the `mindweave` command available globally
+```
+
+It asks for an API key on first launch and saves it to `~/.mindweave/.env`, so every
+project afterwards just works. `mindweave --help` covers the launch flags; everything
+else is configured inside a session.
+
+
+## What it can do
+
+Short version, one line each. The depth is in the linked pages.
+
+- **15 providers, 60 models, one key** — DeepSeek, Anthropic, OpenAI, Gemini, xAI,
+  Mistral, Groq, Cerebras, Qwen, Kimi, GLM, Meta, MiniMax, Tencent, plus every
+  tool-capable model on OpenRouter. Only the driver you
+  use is loaded. Switch with `/provider` and `/model`; remembered per project.
+  [PROVIDERS.md](src/drivers/PROVIDERS.md)
+- **Real tools** — read and edit files, multi-file edits, ripgrep search, a shell with
+  background jobs, and sub-agents. Every edit comes back with the language server's own
+  errors for that file. Read-before-edit is enforced, and `/undo` is a real net.
+- **Deterministic code intelligence** — a background lane indexes your repo with
+  tree-sitter and language servers, costing no tokens, so the agent understands the
+  codebase, not just the open file.
+- **Session memory** — automatic compaction plus a running state summary that survives
+  it, and the agent can read its own earlier sessions in a project.
+- **Project notes** — MINDWEAVE.md is loaded every session and maintained by the agent;
+  `@./path` imports split it, a per-folder one applies only there, and `~/.mindweave/`
+  holds machine-wide notes. `/init` writes the first one.
+- **MCP servers** — connect external tool servers with `/mcp add` or in plain words;
+  their output is untrusted by default. [docs/MCP.md](docs/MCP.md)
+- **Images and web search** — drop a screenshot or write `@shot.png`; ask about a recent
+  release and it looks it up, through your own provider.
+- **Seeing your app** — it can capture one named window (never the whole screen, and it
+  asks first) to tell an app that started from one that works.
+- **Per-project governor** — standing rules, reusable skills, and forbidden paths or
+  commands the agent must respect.
+
+## Using it
+
+| Command | What it does |
+| --- | --- |
+| `/help` | Lists every command |
+| `/init` | Writes MINDWEAVE.md, the project notes loaded every session |
+| `/provider` · `/model` · `/think` | Who answers, which model, how hard it reasons |
+| `/clear` · `/continue` | Start fresh, or resume an earlier session |
+| `/undo` | Reverts what the last turn changed |
+| `/update` | Installs the newest version and reopens on this conversation |
+| `/mcp` | Manages connected MCP servers |
+| `/feedback` | Sends the maintainer a message, no account needed |
+| `shift-tab` | Cycles interaction modes |
+
+Type while it works and your message queues; press up to take it back and edit it.
+
+## Read more
+
+| | |
+| --- | --- |
+| [CHANGELOG.md](CHANGELOG.md) | What changed, and what each fix actually was |
+| [KNOWN-ISSUES.md](KNOWN-ISSUES.md) | What is broken or unfinished, written down rather than carried quietly |
+| [PHILOSOPHY.md](PHILOSOPHY.md) | How the project is run and what gets into the core. Short, and the honest version |
+| [BOUNDARY.md](BOUNDARY.md) | What belongs in the core versus a driver |
+| [CONTRIBUTING.md](CONTRIBUTING.md) | Mechanics, and what is most useful to report |
+| [SECURITY.md](SECURITY.md) | Reporting a vulnerability |
+| [docs/MCP.md](docs/MCP.md) | Connecting and trusting MCP servers |
+
+## Contributing
+
+Contributions are welcome, especially **model drivers**: Ollama is a stub waiting for
+someone, and a driver is a smaller job than people expect, owning one provider's wire
+format and nothing about how the agent behaves.
+
+Small fixes and reproduced bugs with a failing test can go straight to a pull request.
+For anything larger, start a Discussion first. AI-assisted contributions are fine; the
+one rule is that you understand and have tested what you are submitting.
+
+Bug reports are genuinely the most useful thing you can send. Mindweave is developed by
+running it on real projects and fixing what breaks, and nearly everything in the
+changelog started as a failure someone watched happen. [What to include, and what is
+especially worth reporting.](CONTRIBUTING.md#reporting-a-bug)
+
+Questions are welcome at zallinimann@gmail.com
+
+## License
+
+[Apache License 2.0](LICENSE).

package/dist/cli/App.js

@@ -35,7 +35,8 @@ import { appendForbidden, appendForbiddenCommand } from "../governor/write.js";
 import { addRoot, removeRoot } from "../tools/workspace.js";
 import { discoverRelatedRoots } from "../tools/workspaceDiscover.js";
 import { rootLabel, rootsOf, relativize } from "../tools/paths.js";
-import { APPROVAL_DISMISSED, APPROVAL_TEXT } from "../tools/approval.js";
+import { APPROVAL_DISMISSED, APPROVAL_TEXT, readFreeText } from "../tools/approval.js";
+import { buildFeedback, previewOf, refuseReason, sendFeedback, withAddition } from "./feedback.js";
 import { KeySetup } from "./components/KeySetup.js";
 import { setupView } from "./keySetup.js";
 import { KeyManager } from "./components/KeyManager.js";
@@ -1432,7 +1433,7 @@ export function App({ resumeSessionId, initialScreen }) {
     /**
      * Write a server's config and connect it live — shared by the typed `/mcp add`, the
      * `/mcp` box's Add and Edit, and (indirectly, through the same writer) the
-     * `add_mcp_server` tool. One path so a config any of them accepts is one the others
+     * `mcp_server` tool. One path so a config any of them accepts is one the others
      * would too. Connects immediately: writing the file and telling the user to restart
      * would defeat the entire point of a guided add.
      */
@@ -1481,7 +1482,7 @@ export function App({ resumeSessionId, initialScreen }) {
     /**
      * `/mcp add <name> <command|url> [args…]` and `/mcp remove <name>`.
      *
-     * Shares its parser and writer with the `add_mcp_server` tool, so a config the command
+     * Shares its parser and writer with the `mcp_server` tool, so a config the command
      * accepts is exactly one the tool would, and vice versa.
      */
     async function mcpConfigCommand(arg) {
@@ -2320,7 +2321,7 @@ export function App({ resumeSessionId, initialScreen }) {
     }
     // Hand a free-text directive to the model with an instruction wrapper, as its own
     // turn — used by the manual /rules and /skills commands (the model does the
-    // "rewrite it well and save it" work via remember_rule / create_skill).
+    // "rewrite it well and save it" work via remember_rule / the skill tool).
     async function runDirective(instruction, activity) {
         const s = session.current;
         if (!s)
@@ -2422,6 +2423,41 @@ export function App({ resumeSessionId, initialScreen }) {
         }
         // /analytics — the anonymous usage ping. Bare opens the on/off switch, the same
         // fixed box every other setting uses; an argument acts directly, same as /model.
+        // /feedback — a message to the maintainer, from inside the app. No account, no issue
+        // tracker, no mail client: the whole point is that it costs the sender nothing. What
+        // leaves the machine is decided in cli/feedback.ts and shown here before it goes.
+        if (name === "/feedback") {
+            const typed = arg.trim();
+            const message = typed || (await askApproval.current("What would you like to tell the maintainer?", ["Cancel"], undefined, undefined, { label: "Your message", placeholder: "what is missing, broken, or worth adding" }).then((answer) => readFreeText(answer) ?? ""));
+            if (!message.trim())
+                return;
+            // The confirm LOOPS, because the moment you read your own message back is the
+            // moment you remember the thing you left out — usually an email address to reply
+            // to. Typing in the box's own text row adds a line and shows the message again,
+            // rather than cancelling and retyping the whole thing.
+            let text = message;
+            for (;;) {
+                const refusal = refuseReason(text);
+                if (refusal)
+                    return say(refusal);
+                const feedback = buildFeedback(text, appVersion());
+                const choice = await askApproval.current(previewOf(feedback), ["Send it", "Don't send"], undefined, undefined, {
+                    label: "add a line",
+                    placeholder: "your email for a reply, or more detail",
+                });
+                const added = readFreeText(choice);
+                if (added !== null) {
+                    text = withAddition(text, added);
+                    continue;
+                }
+                if (choice !== "Send it")
+                    return note("Not sent.");
+                note("sending…");
+                const result = await sendFeedback(feedback);
+                note(result.message, result.ok ? undefined : { error: true });
+                return;
+            }
+        }
         if (name === "/analytics") {
             const verb = arg.trim().toLowerCase();
             if (verb === "on" || verb === "off") {
@@ -2674,7 +2710,7 @@ export function App({ resumeSessionId, initialScreen }) {
                 // `/model openrouter` alone: that provider's picker.
                 if (!words)
                     return setOverlay({ kind: "model", providerId });
-                const picked = providerId === current ? here : resolveChoice(words, orderedModelsOf(providerId), "model");
+                const picked = providerId === current && words === arg.trim() ? here : resolveChoice(words, orderedModelsOf(providerId), "model");
                 if (picked.kind === "error")
                     return say(picked.message);
                 // Several fit: show exactly those, in the picker, rather than a list to retype from.
@@ -2862,7 +2898,7 @@ export function App({ resumeSessionId, initialScreen }) {
         if (name === "/skills") {
             if (arg) {
                 await runDirective(`The user wants a new reusable skill for this project: "${arg}". Design it and save it ` +
-                    `with create_skill — a short invocation name, a one-line description, and the procedure ` +
+                    `with the skill tool — a short invocation name, a one-line description, and the procedure ` +
                     `as a clear markdown checklist (use $ARGUMENTS/$1 if it should take input). Then confirm.`, "creating a skill…");
                 return;
             }
@@ -2999,7 +3035,7 @@ export function App({ resumeSessionId, initialScreen }) {
         return (_jsx(TrustGate, { rows: rows, cwd: startCwd.current, breadth: trustBreadth, warning: breadthWarning(trustBreadth, startCwd.current), persists: trustPersists(trustBreadth), version: versionLabel(), docsUrl: MINDWEAVE_DOCS_URL, onTrust: () => {
                 rememberTrust(projectDir(startCwd.current), trustBreadth);
                 setTrustOpen(false);
-            }, onQuit: () => exit() }));
+            }, onQuit: () => process.exit(0) }));
     }
     // FIRST RUN: nothing can run yet, so offer every provider rather than one. The user
     // adds as many keys as they like and continues when they are ready. See KeySetup.
@@ -3151,9 +3187,12 @@ export function App({ resumeSessionId, initialScreen }) {
                 return {
                     label: m.label + (m.id === id ? "  ✓" : ""),
                     description: m.description ? `${facts} · ${m.description}` : facts,
+                    // Searched but not shown, so a filter opened by `/model <words>` matches the same
+                    // things the words matched: the id as well as the name.
+                    keywords: m.id,
                 };
             });
-            return (_jsx(Picker, { title: `Choose a ${pickerProvider} model`, items: items, width: width, maxRows: maxRows, initialIndex: Math.max(0, models.findIndex((m) => m.id === id)), initialFilter: overlay.filter, onSelect: onOverlaySelect, onCancel: onOverlayCancel }));
+            return (_jsx(Picker, { title: `Choose ${/^[aeiou]/i.test(pickerProvider) ? "an" : "a"} ${pickerProvider} model`, items: items, width: width, maxRows: maxRows, initialIndex: Math.max(0, models.findIndex((m) => m.id === id)), initialFilter: overlay.filter, onSelect: onOverlaySelect, onCancel: onOverlayCancel }));
         }
         if (overlay.kind === "think") {
             const model = cur?.modelConfig.model ?? DEFAULT_MODEL_CONFIG.model;