code-muse 0.0.1__py3-none-any.whl

code_muse/plugins/mindpack/tools.py

@@ -0,0 +1,387 @@
+"""MindPack tools — ask_mindpack tool and orchestrator wiring.
+
+Initialises a singleton MindPackOrchestrator with the five default
+experts and registers the advisory deep-solve tool via the plugin
+hook system.  Data models live in ``schemas.py`` to avoid circular
+imports between this module and ``orchestration.py``.
+
+Nested-workflow safety (Epic 5 & 6):
+- ``MindPackInvocationContext`` tracks per-prompt call counts and depth.
+- ``ask_mindpack`` cannot be called recursively (max_depth defaults to 1).
+- ``max_calls_per_prompt`` prevents execution loops.
+"""
+
+import logging
+from contextvars import ContextVar
+from dataclasses import dataclass, field
+from typing import Literal
+
+from pydantic_ai import RunContext
+
+from code_muse.plugins.mindpack.orchestration import MindPackOrchestrator
+from code_muse.plugins.mindpack.schemas import (
+    AskMindPackInput,
+    AskMindPackOutput,
+    ExpertDescriptor,
+    MindPackNestedConfig,
+    MindPackRankedOption,
+)
+
+# Backward-compatible re-exports — external code importing from
+# ``tools`` still works, but canonical location is ``schemas``.
+__all__ = [
+    "AskMindPackInput",
+    "AskMindPackOutput",
+    "ExpertDescriptor",
+    "MindPackRankedOption",
+    "MindPackInvocationContext",
+    "register_ask_mindpack",
+]
+
+logger = logging.getLogger(__name__)
+
+# ---------------------------------------------------------------------------
+# INI config helpers (local copy to avoid cross-module import cycles)
+# ---------------------------------------------------------------------------
+
+
+def _get_config_str(key: str, default: str) -> str:
+    from code_muse.config import get_value
+
+    return get_value(key) or default
+
+
+def _get_config_int(key: str, default: int) -> int:
+    from code_muse.config import get_value
+
+    val = get_value(key)
+    if val is None:
+        return default
+    try:
+        return int(val)
+    except ValueError, TypeError:
+        return default
+
+
+def _get_config_bool(key: str, default: bool) -> bool:
+    from code_muse.config import get_value
+
+    val = get_value(key)
+    if val is None:
+        return default
+    return str(val).lower() in ("1", "true", "yes", "on")
+
+
+def load_nested_config_from_ini() -> MindPackNestedConfig:
+    """Build a ``MindPackNestedConfig`` from INI settings.
+
+    Reads ``packmind_nested_enabled``, ``packmind_nested_max_depth``,
+    ``packmind_nested_max_calls_per_prompt``, and
+    ``packmind_nested_timeout_sec`` from the Muse INI config.
+    """
+    return MindPackNestedConfig(
+        enabled=_get_config_bool("packmind_nested_enabled", True),
+        max_depth=_get_config_int("packmind_nested_max_depth", 1),
+        max_calls_per_prompt=_get_config_int("packmind_nested_max_calls_per_prompt", 2),
+        timeout_sec=_get_config_int("packmind_nested_timeout_sec", 90),
+    )
+
+
+# ---------------------------------------------------------------------------
+# MindPackInvocationContext — nested-workflow guardrails
+# ---------------------------------------------------------------------------
+
+
+@dataclass
+class MindPackInvocationContext:
+    """Runtime context for a single MindPack invocation.
+
+    Tracks nesting depth (to prevent recursive ``ask_mindpack`` calls)
+    and the number of invocations already made in the current prompt
+    (to enforce ``max_calls_per_prompt``).
+    """
+
+    nested_depth: int = 0
+    max_depth: int = 1
+    max_calls_per_prompt: int = 2
+    session_id: str | None = None
+    call_count: int = field(default=0, repr=False)
+
+    def is_recursion_blocked(self) -> bool:
+        """True if we are already inside a MindPack call (depth ≥ 1)."""
+        return self.nested_depth >= self.max_depth
+
+    def is_call_limit_reached(self) -> bool:
+        """True if the per-prompt call cap has been exceeded."""
+        return self.call_count >= self.max_calls_per_prompt
+
+    def increment_call(self) -> None:
+        self.call_count += 1
+        self.nested_depth += 1
+
+
+# ContextVar tracking the *current* invocation (None when outside ask_mindpack)
+_mindpack_ctx_var: ContextVar[MindPackInvocationContext | None] = ContextVar(
+    "mindpack_ctx", default=None
+)
+
+# ContextVar tracking how many times ask_mindpack has been called in this prompt
+_mindpack_call_count_var: ContextVar[int] = ContextVar("mindpack_call_count", default=0)
+
+
+def _check_and_increment_call(
+    nested_config: MindPackNestedConfig,
+) -> MindPackInvocationContext:
+    """Validate recursion / call-count guardrails and return a fresh context.
+
+    Raises:
+        RuntimeError: with an explicit rejection message if recursion limits
+        or the per-prompt call cap are exceeded.
+    """
+    # 1. Check if we are already inside a MindPack invocation (recursion)
+    existing_ctx = _mindpack_ctx_var.get()
+    if existing_ctx is not None:
+        raise RuntimeError(
+            "❌ MindPack recursion blocked: ask_mindpack cannot be called recursively. "
+            f"Current nested_depth={existing_ctx.nested_depth}, max_depth={existing_ctx.max_depth}. "
+            "Experts are already running in advisory mode; invoking another MindPack panel "
+            "inside an active consultation would create an infinite loop."
+        )
+
+    # 2. Check per-prompt call count
+    current_calls = _mindpack_call_count_var.get()
+    if current_calls >= nested_config.max_calls_per_prompt:
+        raise RuntimeError(
+            "❌ MindPack call-limit reached: ask_mindpack has already been invoked "
+            f"{current_calls} time(s) in this prompt (max_calls_per_prompt="
+            f"{nested_config.max_calls_per_prompt}). To prevent execution loops, "
+            "further MindPack calls are blocked for the remainder of this turn."
+        )
+
+    # 3. Increment counters and build a new context
+    call_token = _mindpack_call_count_var.set(current_calls + 1)
+    new_ctx = MindPackInvocationContext(
+        nested_depth=1,
+        max_depth=nested_config.max_depth,
+        max_calls_per_prompt=nested_config.max_calls_per_prompt,
+        call_count=current_calls + 1,
+    )
+    ctx_token = _mindpack_ctx_var.set(new_ctx)
+
+    # Return tokens so the caller can reset them in a finally block
+    new_ctx._call_token = call_token  # type: ignore[attr-defined]
+    new_ctx._ctx_token = ctx_token  # type: ignore[attr-defined]
+    return new_ctx
+
+
+def _reset_context(ctx: MindPackInvocationContext) -> None:
+    """Reset ContextVars after a MindPack invocation finishes."""
+    try:
+        _mindpack_ctx_var.reset(ctx._ctx_token)  # type: ignore[attr-defined]
+    except Exception:
+        pass
+    try:
+        _mindpack_call_count_var.reset(ctx._call_token)  # type: ignore[attr-defined]
+    except Exception:
+        pass
+
+
+# ---------------------------------------------------------------------------
+# Singleton orchestrator + default expert registration
+# ---------------------------------------------------------------------------
+
+orchestrator = MindPackOrchestrator()
+
+_DEFAULT_EXPERTS: list[ExpertDescriptor] = [
+    ExpertDescriptor(
+        name="Scout",
+        speciality="codebase exploration & context gathering",
+        model="fast",
+        system_prompt_fragment=(
+            "You are Scout, the codebase explorer. Your job is to quickly "
+            "scan relevant files, trace dependencies, and surface the "
+            "context needed to understand a problem. Focus on what exists "
+            "and how it connects."
+        ),
+    ),
+    ExpertDescriptor(
+        name="Architect",
+        speciality="design & architecture decisions",
+        model="strong",
+        system_prompt_fragment=(
+            "You are Architect, the design thinker. Your job is to "
+            "propose clean structural solutions, weigh trade-offs, "
+            "and ensure the plan fits the existing codebase architecture."
+        ),
+    ),
+    ExpertDescriptor(
+        name="Strategic Architect",
+        speciality="big-picture architecture & long-term vision",
+        model="strong",
+        system_prompt_fragment=(
+            "You are Strategic Architect, the big-picture thinker. Your job is to "
+            "step back and design the long-term architecture — system boundaries, "
+            "module ownership, technology choices, and how the plan fits the "
+            "project's 6-month roadmap. Think in terms of maintainability, "
+            "scalability, and technical debt avoidance. Propose the ideal end-state "
+            "and work backwards to a phased delivery plan."
+        ),
+    ),
+    ExpertDescriptor(
+        name="Systems Architect",
+        speciality="component integration & data flow design",
+        model="strong",
+        system_prompt_fragment=(
+            "You are Systems Architect, the integration specialist. Your job is to "
+            "design how components communicate — API contracts, data models, event "
+            "flows, and service boundaries. Focus on the seams between modules: "
+            "what data moves where, what interfaces need to change, and how to "
+            "minimize coupling. Your plans should make the integration surface "
+            "explicit and testable."
+        ),
+    ),
+    ExpertDescriptor(
+        name="Pragmatic Architect",
+        speciality="incremental delivery & risk-aware planning",
+        model="strong",
+        system_prompt_fragment=(
+            "You are Pragmatic Architect, the delivery-focused planner. Your job is "
+            "to find the smallest viable change that moves the needle. You balance "
+            "idealism with reality — prefer incremental refactors over rewrites, "
+            "identify what can be deferred, and always have a rollback strategy. "
+            "Your plans prioritize developer velocity, reversibility, and "
+            "shipping value early."
+        ),
+    ),
+    ExpertDescriptor(
+        name="Watchdog",
+        speciality="risk assessment & edge-case analysis",
+        model="medium",
+        system_prompt_fragment=(
+            "You are Watchdog, the risk analyst. Your job is to "
+            "identify failure modes, security concerns, edge cases, "
+            "and anything that could go wrong with a proposed plan."
+        ),
+    ),
+    ExpertDescriptor(
+        name="Test Planner",
+        speciality="validation strategy & test design",
+        model="medium",
+        system_prompt_fragment=(
+            "You are Test Planner, the validation strategist. Your job is "
+            "to design the smallest effective test suite that confirms "
+            "the plan works and catches regressions."
+        ),
+    ),
+    ExpertDescriptor(
+        name="Challenger",
+        speciality="adversarial review & assumption questioning",
+        model="strong",
+        system_prompt_fragment=(
+            "You are Challenger, the adversarial reviewer. Your job is "
+            "to poke holes in every proposal, question assumptions, "
+            "and argue for simpler alternatives when they exist."
+        ),
+    ),
+]
+
+orchestrator.register_experts(_DEFAULT_EXPERTS)
+orchestrator.load_experts()  # Load any custom experts from disk
+orchestrator.load_profiles()  # Load profiles (seeds defaults on first run)
+
+logger.info(
+    "MindPack orchestrator initialised with %d default expert(s): %s",
+    len(_DEFAULT_EXPERTS),
+    [e.name for e in _DEFAULT_EXPERTS],
+)
+
+# ---------------------------------------------------------------------------
+# Tool registration
+# ---------------------------------------------------------------------------
+
+
+def register_ask_mindpack(agent):
+    """Register the ask_mindpack tool on the given agent."""
+
+    @agent.tool
+    async def ask_mindpack(
+        context: RunContext,
+        problem_statement: str,
+        current_goal: str,
+        current_plan: str | None = None,
+        what_has_been_tried: list[str] | None = None,
+        relevant_files: list[str] | None = None,
+        observed_errors: list[str] | None = None,
+        uncertainty: str | None = None,
+        desired_output: Literal[
+            "plan",
+            "review",
+            "debug_strategy",
+            "architecture_decision",
+            "test_strategy",
+            "compare_options",
+        ] = "plan",
+        max_experts: int | None = None,
+    ) -> AskMindPackOutput:
+        """Request a deep advisory analysis from the MindPack expert pool.
+
+        Use this tool when you encounter a problem that needs deeper
+        reasoning, multiple expert perspectives, or comparison of
+        approaches.  MindPack runs in nested advisory mode: experts
+        are read-only, make no file edits, and cannot call ask_mindpack
+        recursively.
+
+        Do NOT call this for simple edits or direct user requests
+        that require only a single obvious action.
+
+        Args:
+            problem_statement: The specific problem to solve.
+            current_goal: What you are currently trying to achieve.
+            current_plan: Your current plan, if any.
+            what_has_been_tried: Approaches already attempted.
+            relevant_files: File paths relevant to the problem.
+            observed_errors: Error messages or unexpected behaviours.
+            uncertainty: What you are uncertain about.
+            desired_output: Kind of advisory output you want.
+            max_experts: Cap on number of experts to consult.
+        """
+        nested_config = load_nested_config_from_ini()
+
+        # Epic 5 & 6 — enforce nested-workflow guardrails
+        invocation_ctx = _check_and_increment_call(nested_config)
+        try:
+            logger.info(
+                "ask_mindpack invoked: desired_output=%s, problem=%s "
+                "depth=%d/%d calls=%d/%d",
+                desired_output,
+                problem_statement[:120],
+                invocation_ctx.nested_depth,
+                invocation_ctx.max_depth,
+                invocation_ctx.call_count,
+                invocation_ctx.max_calls_per_prompt,
+            )
+
+            request = AskMindPackInput(
+                problem_statement=problem_statement,
+                current_goal=current_goal,
+                current_plan=current_plan,
+                what_has_been_tried=what_has_been_tried or [],
+                relevant_files=relevant_files or [],
+                observed_errors=observed_errors or [],
+                uncertainty=uncertainty,
+                desired_output=desired_output,
+                max_experts=max_experts,
+            )
+
+            # Run orchestration with invocation context so the orchestrator
+            # knows it is inside a nested workflow (e.g., for timeout caps).
+            raw_output = await orchestrator.consult(
+                request,
+                invocation_context=invocation_ctx,
+                nested_config=nested_config,
+            )
+            return raw_output
+        finally:
+            _reset_context(invocation_ctx)
+
+    return ask_mindpack

code_muse/plugins/oauth_muse_html.py

@@ -0,0 +1,226 @@
+"""Shared HTML templates for OAuth result pages with clean colored status indicators."""
+
+CLAUDE_LOGO_URL = "https://voideditor.com/claude-icon.png"
+CHATGPT_LOGO_URL = (
+    "https://freelogopng.com/images/all_img/1681038325chatgpt-logo-transparent.png"
+)
+GEMINI_LOGO_URL = "https://upload.wikimedia.org/wikipedia/commons/thumb/8/8a/Google_Gemini_logo.svg/512px-Google_Gemini_logo.svg.png"
+
+
+def oauth_success_html(service_name: str, extra_message: str | None = None) -> str:
+    """Return a clean OAuth success HTML page with colored status indicators."""
+    clean_service = service_name.strip() or "OAuth"
+    detail = f"<p class='detail'>{extra_message}</p>" if extra_message else ""
+    projectile, rival_url, rival_alt, target_modifier = _service_targets(clean_service)
+    target_classes = "target" if not target_modifier else f"target {target_modifier}"
+    return (
+        "<!DOCTYPE html>"
+        "<html lang='en'><head><meta charset='utf-8'>"
+        "<title>Auth Complete</title>"
+        "<style>"
+        "html,body{margin:0;padding:0;height:100%;overflow:hidden;font-family:'Inter',-apple-system,BlinkMacSystemFont,'Segoe UI',sans-serif;background:linear-gradient(135deg,#0f172a 0%,#111827 45%,#1f2937 100%);color:#e5e7eb;}"
+        "body{display:flex;align-items:center;justify-content:center;}"
+        ".panel{position:relative;width:90%;max-width:880px;padding:60px;background:rgba(15,23,42,0.72);border-radius:32px;backdrop-filter:blur(14px);box-shadow:0 30px 90px rgba(8,11,18,0.7);text-align:center;border:1px solid rgba(148,163,184,0.25);}"
+        "h1{font-size:3.4em;margin:0;color:#f1f5f9;text-shadow:0 14px 40px rgba(8,11,18,0.55);letter-spacing:1px;}"
+        "p{font-size:1.25em;margin:16px 0;color:#cbd5f5;}"
+        ".detail{font-size:1.1em;opacity:0.9;}"
+        ".mega{display:inline-block;font-size:1.35em;margin-top:14px;color:#f97316;}"
+        ".confetti{position:absolute;top:50%;left:50%;transform:translate(-50%,-50%);width:200%;height:200%;pointer-events:none;mix-blend-mode:screen;}"
+        ".confetti span{position:absolute;font-size:3.4em;animation:floaty 6s ease-in-out infinite;}"
+        ".confetti span.ok{color:#4ade80;}"
+        ".confetti span.warn{color:#facc15;}"
+        ".confetti span.info{color:#60a5fa;}"
+        "@keyframes floaty{0%,100%{transform:translate3d(0,0,0) rotate(0deg);}35%{transform:translate3d(0,-70px,0) rotate(10deg);}65%{transform:translate3d(0,-90px,0) rotate(-12deg);}90%{transform:translate3d(0,-60px,0) rotate(6deg);}}"
+        ".confetti span:nth-child(odd){animation-duration:7.2s;}"
+        ".confetti span:nth-child(3n){animation-duration:8.6s;}"
+        ".confetti span:nth-child(4n){animation-duration:5.9s;}"
+        ".confetti span:nth-child(5n){animation-duration:9.4s;}"
+        ".artillery{position:absolute;bottom:12%;left:0;width:100%;max-width:1100px;height:240px;pointer-events:none;overflow:visible;}"
+        ".artillery .cannon{position:absolute;bottom:0;font-size:3.4em;filter:drop-shadow(0 12px 32px rgba(249,115,22,0.45));}"
+        ".artillery .cannon.left{left:4%;color:#4ade80;}"
+        ".artillery .cannon.right{left:12%;transform:rotate(-4deg);color:#f87171;}"
+        ".artillery .shell{position:absolute;left:10%;font-size:2.6em;animation:strafe 2.6s ease-out infinite;color:#facc15;text-shadow:0 0 14px rgba(250,204,21,0.45);}"
+        "@keyframes strafe{0%{left:10%;opacity:1;}60%{left:72%;opacity:1;}100%{left:82%;opacity:0;}}"
+        ".target{position:absolute;top:175px;right:-10%;width:220px;filter:drop-shadow(0 24px 46px rgba(8,11,18,0.72));animation:targetShake 1.9s ease-in-out infinite;}"
+        ".target img{width:200px;height:auto;border-radius:18px;background:#0f172a;padding:16px;border:1px solid rgba(148,163,184,0.35);}"
+        ".target.invert img{filter:brightness(1.2) saturate(1.15);background:rgba(15,23,42,0.9);}"
+        "@keyframes targetShake{0%,100%{transform:rotate(0deg) scale(1);}30%{transform:rotate(-4deg) scale(1.05);}60%{transform:rotate(3deg) scale(0.97);}85%{transform:rotate(-2deg) scale(1.04);}}"
+        ".target::after{content:'';position:absolute;top:50%;left:50%;width:220px;height:220px;border-radius:50%;background:radial-gradient(circle,rgba(74,222,128,0.35)0%,rgba(74,222,128,0)70%);transform:translate(-50%,-50%) scale(0);animation:impact 2.6s ease-out infinite;opacity:0;mix-blend-mode:screen;}"
+        "@keyframes impact{0%,60%{transform:translate(-50%,-50%) scale(0);opacity:0;}70%{transform:translate(-50%,-50%) scale(1.2);opacity:1;}100%{transform:translate(-50%,-50%) scale(1.5);opacity:0;}}"
+        "</style>"
+        "</head><body>"
+        "<div class='panel'>"
+        "<div class='confetti'>"
+        + "".join(
+            f"<span class='{cls}' style='left:{left}%;top:{top}%;animation-delay:{delay}s;'>{token}</span>"
+            for left, top, delay, token, cls in _SUCCESS_TOKENS
+        )
+        + "</div>"
+        f"<h1>[Done] {clean_service} Auth Complete</h1>"
+        "<p class='mega'>Authentication complete. Token delivered.</p>"
+        f"{detail}"
+        f"<p>Token payload delivered to {rival_alt}.</p>"
+        "<p>This window will auto-close shortly.</p>"
+        "<p class='mega'>Return to your session.</p>"
+        f"<div class='{target_classes}'><img src='{rival_url}' alt='{rival_alt}'></div>"
+        "<div class='artillery'>" + _build_artillery(projectile) + "</div>"
+        "</div>"
+        "<script>setTimeout(()=>window.close(),3500);</script>"
+        "</body></html>"
+    )
+
+
+def oauth_failure_html(service_name: str, reason: str) -> str:
+    """Return a clean OAuth failure HTML page with colored status indicators."""
+    clean_service = service_name.strip() or "OAuth"
+    clean_reason = reason.strip() or "Authentication failed"
+    projectile, rival_url, rival_alt, target_modifier = _service_targets(clean_service)
+    target_classes = "target" if not target_modifier else f"target {target_modifier}"
+    return (
+        "<!DOCTYPE html>"
+        "<html lang='en'><head><meta charset='utf-8'>"
+        "<title>Auth Failed</title>"
+        "<style>"
+        "html,body{margin:0;padding:0;height:100%;overflow:hidden;font-family:'Inter',-apple-system,BlinkMacSystemFont,'Segoe UI',sans-serif;background:linear-gradient(160deg,#101827 0%,#0b1120 100%);color:#e2e8f0;}"
+        "body{display:flex;align-items:center;justify-content:center;}"
+        ".panel{position:relative;width:90%;max-width:780px;padding:55px;background:rgba(10,13,23,0.78);border-radius:30px;box-shadow:0 26px 80px rgba(2,6,23,0.78);text-align:center;border:1px solid rgba(71,85,105,0.35);}"
+        "h1{font-size:3em;margin:0 0 14px;text-shadow:0 16px 36px rgba(15,23,42,0.7);color:#f87171;}"
+        "p{font-size:1.2em;margin:14px 0;line-height:1.6;color:#cbd5f5;}"
+        ".alert{font-size:1.35em;margin:18px 0;color:#fda4af;}"
+        ".tearstorm{position:absolute;top:50%;left:50%;transform:translate(-50%,-50%);width:190%;height:190%;pointer-events:none;mix-blend-mode:screen;}"
+        ".tearstorm span{position:absolute;font-size:3.2em;animation:weep 4.8s ease-in-out infinite;}"
+        ".tearstorm span.fail{color:#f87171;}"
+        ".tearstorm span.warn{color:#fbbf24;}"
+        ".tearstorm span.info{color:#60a5fa;}"
+        "@keyframes weep{0%{transform:translate3d(0,-10px,0) rotate(-6deg);opacity:0.85;}35%{transform:translate3d(-20px,18px,0) rotate(8deg);opacity:1;}65%{transform:translate3d(25px,28px,0) rotate(-12deg);opacity:0.8;}100%{transform:translate3d(0,60px,0) rotate(0deg);opacity:0;}}"
+        ".tearstorm span:nth-child(odd){animation-duration:5.8s;}"
+        ".tearstorm span:nth-child(3n){animation-duration:6.4s;}"
+        ".tearstorm span:nth-child(4n){animation-duration:7.3s;}"
+        ".buttons{margin-top:26px;}"
+        ".buttons a{display:inline-block;margin:6px 12px;padding:12px 28px;border-radius:999px;background:rgba(59,130,246,0.16);color:#bfdbfe;text-decoration:none;font-weight:600;border:1px solid rgba(96,165,250,0.4);transition:all 0.3s;}"
+        ".buttons a:hover{background:rgba(96,165,250,0.28);transform:translateY(-2px);}"
+        ".battlefield{position:absolute;bottom:-25px;left:0;width:100%;max-width:960px;height:220px;pointer-events:none;}"
+        ".battlefield .shell{position:absolute;left:10%;font-size:2.4em;color:#38bdf8;text-shadow:0 0 12px rgba(56,189,248,0.45);animation:strafeSad 3s ease-out infinite;}"
+        "@keyframes strafeSad{0%{left:10%;opacity:1;}65%{left:70%;opacity:1;}100%{left:80%;opacity:0;}}"
+        ".battlefield .target{position:absolute;top:16px;right:6%;width:220px;filter:drop-shadow(0 20px 44px rgba(2,6,23,0.78));animation:sway 2s ease-in-out infinite;}"
+        ".battlefield .target img{width:200px;height:auto;border-radius:18px;background:#0b1120;padding:16px;border:1px solid rgba(96,165,250,0.4);}"
+        ".battlefield .target.invert img{filter:brightness(1.2) saturate(1.15);background:rgba(15,23,42,0.9);}"
+        "@keyframes sway{0%,100%{transform:rotate(0deg);}40%{transform:rotate(-6deg);}70%{transform:rotate(5deg);}}"
+        "</style>"
+        "</head><body>"
+        "<div class='panel'>"
+        "<div class='tearstorm'>"
+        + "".join(
+            f"<span class='{cls}' style='left:{left}%;top:{top}%;animation-delay:{delay}s;'>{token}</span>"
+            for left, top, delay, token, cls in _FAILURE_TOKENS
+        )
+        + "</div>"
+        f"<h1>[Fail] {clean_service} Auth Failed</h1>"
+        "<p class='alert'>Authentication failed.</p>"
+        f"<p>{clean_reason}</p>"
+        "<p>Retry from Muse.</p>"
+        f"<p>Re-attempt authentication with {rival_alt}.</p>"
+        "<div class='buttons'>"
+        "<a href='https://muse.dev' target='_blank'>Documentation</a>"
+        "<a href='https://github.com/asx8678/muse' target='_blank'>GitHub</a>"
+        "</div>"
+        "<div class='battlefield'>"
+        + _build_artillery(projectile, shells_only=True)
+        + f"<div class='{target_classes}'><img src='{rival_url}' alt='{rival_alt}'></div>"
+        + "</div>"
+        "</div>"
+        "</body></html>"
+    )
+
+
+_SUCCESS_TOKENS = (
+    (5, 12, 0.0, "[Done]", "ok"),
+    (18, 28, 0.2, "[Done]", "ok"),
+    (32, 6, 1.1, "[OK]", "ok"),
+    (46, 18, 0.5, "[Done]", "ok"),
+    (62, 9, 0.8, "[OK]", "ok"),
+    (76, 22, 1.3, "[Done]", "ok"),
+    (88, 14, 0.4, "[OK]", "ok"),
+    (12, 48, 0.6, "[Run]", "info"),
+    (28, 58, 1.7, "[>>]", "info"),
+    (44, 42, 0.9, "[Run]", "info"),
+    (58, 52, 1.5, "[>>]", "info"),
+    (72, 46, 0.3, "[Run]", "info"),
+    (86, 54, 1.1, "[>>]", "info"),
+    (8, 72, 0.7, "[Done]", "ok"),
+    (24, 80, 1.2, "[OK]", "ok"),
+    (40, 74, 0.2, "[Done]", "ok"),
+    (56, 66, 1.6, "[Run]", "info"),
+    (70, 78, 1.0, "[>>]", "info"),
+    (84, 70, 1.4, "[Run]", "info"),
+    (16, 90, 0.5, "[Done]", "ok"),
+    (32, 92, 1.9, "[OK]", "ok"),
+    (48, 88, 1.1, "[Run]", "info"),
+    (64, 94, 1.8, "[>>]", "info"),
+    (78, 88, 0.6, "[Done]", "ok"),
+    (90, 82, 1.3, "[OK]", "ok"),
+)
+
+_FAILURE_TOKENS = (
+    (8, 6, 0.0, "[Fail]", "fail"),
+    (22, 18, 0.3, "[!!]", "fail"),
+    (36, 10, 0.6, "[Fail]", "fail"),
+    (50, 20, 0.9, "[!!]", "fail"),
+    (64, 8, 1.2, "[Fail]", "fail"),
+    (78, 16, 1.5, "[!!]", "fail"),
+    (12, 38, 0.4, "[Warn]", "warn"),
+    (28, 44, 0.7, "[??]", "warn"),
+    (42, 34, 1.0, "[Warn]", "warn"),
+    (58, 46, 1.3, "[??]", "warn"),
+    (72, 36, 1.6, "[Warn]", "warn"),
+    (86, 40, 1.9, "[??]", "warn"),
+    (16, 64, 0.5, "[Fail]", "fail"),
+    (32, 70, 0.8, "[!!]", "fail"),
+    (48, 60, 1.1, "[Fail]", "fail"),
+    (62, 74, 1.4, "[!!]", "fail"),
+    (78, 68, 1.7, "[Fail]", "fail"),
+    (90, 72, 2.0, "[!!]", "fail"),
+    (20, 88, 0.6, "[Warn]", "warn"),
+    (36, 92, 0.9, "[??]", "warn"),
+    (52, 86, 1.2, "[Warn]", "warn"),
+    (68, 94, 1.5, "[??]", "warn"),
+    (82, 90, 1.8, "[Warn]", "warn"),
+)
+
+_STRAFE_SHELLS: tuple[tuple[float, float], ...] = (
+    (22.0, 0.0),
+    (28.0, 0.35),
+    (34.0, 0.7),
+    (26.0, 0.2),
+    (32.0, 0.55),
+    (24.0, 0.9),
+    (30.0, 1.25),
+)
+
+
+def _build_artillery(projectile: str, *, shells_only: bool = False) -> str:
+    """Return HTML spans for artillery shells (and cannons when desired)."""
+    shell_markup = []
+    for index, (top, delay) in enumerate(_STRAFE_SHELLS):
+        duration = 2.3 + (index % 3) * 0.25
+        shell_markup.append(
+            f"<span class='shell' style='top:{top}%;animation-delay:-{delay}s;animation-duration:{duration}s;'>{projectile}</span>"
+        )
+    shells = "".join(shell_markup)
+    if shells_only:
+        return shells
+
+    cannons = "<span class='cannon left'>[Launch]</span><span class='cannon right'>[Fire]</span>"
+    return cannons + shells
+
+
+def _service_targets(service_name: str) -> tuple[str, str, str, str]:
+    """Map service names to projectile label and rival logo metadata."""
+    normalized = service_name.lower()
+    if "anthropic" in normalized or "claude" in normalized:
+        return "[Launch]", CLAUDE_LOGO_URL, "Claude logo", ""
+    if "chat" in normalized or "gpt" in normalized:
+        return "[Fire]", CHATGPT_LOGO_URL, "ChatGPT logo", "invert"
+    if "gemini" in normalized or "google" in normalized:
+        return "[Auth]", GEMINI_LOGO_URL, "Gemini logo", ""
+    return "[Fire]", CHATGPT_LOGO_URL, "service logo", "invert"

code_muse/plugins/ollama_setup/__init__.py

@@ -0,0 +1,5 @@
+"""Ollama cloud model setup plugin.
+
+Provides `/ollama-setup <model>` to pull an Ollama cloud model and register
+it in ~/.muse/extra_models.json so it's immediately usable.
+"""

code_muse/plugins/ollama_setup/completer.py

@@ -0,0 +1,36 @@
+"""Tab-completion for ``/ollama-setup <model>``."""
+
+from collections.abc import Iterable
+
+from prompt_toolkit.completion import Completer, Completion
+from prompt_toolkit.document import Document
+
+from code_muse.plugins.ollama_setup.register_callbacks import CLOUD_MODELS
+
+
+class OllamaSetupCompleter(Completer):
+    """Completes cloud model names after ``/ollama-setup ``."""
+
+    trigger = "/ollama-setup"
+
+    def get_completions(
+        self, document: Document, complete_event: object
+    ) -> Iterable[Completion]:
+        text = document.text_before_cursor.lstrip()
+
+        if not text.startswith(self.trigger + " "):
+            return
+
+        # Everything after "/ollama-setup "
+        after = text[len(self.trigger) + 1 :]
+        partial = after.strip().lower()
+        start_pos = -len(after) if after else 0
+
+        for tag, meta in CLOUD_MODELS.items():
+            if not partial or tag.lower().startswith(partial):
+                yield Completion(
+                    tag,
+                    start_position=start_pos,
+                    display=tag,
+                    display_meta=meta["description"],
+                )