@thispointon/kondi-chat 0.1.2

package/src/council-engine/pipeline/test-detect.ts

@@ -0,0 +1,82 @@
+/**
+ * Test Command Auto-Detection
+ * Scans project files to determine the appropriate test command.
+ * Accepts an optional readFile callback to abstract file I/O (Tauri / Node.js).
+ */
+
+export type ReadFileFn = (path: string) => Promise<string | null>;
+
+export interface DetectedTest {
+  command: string;
+  framework: string;
+  confidence: 'high' | 'medium' | 'low';
+}
+
+/**
+ * Detect the appropriate test command for a project directory.
+ * Checks common project files in priority order.
+ *
+ * @param workingDir  Absolute path to the project root
+ * @param readFile    Optional callback that reads a file and returns its content
+ *                    (or null if not found). When omitted, detection is skipped.
+ */
+export async function detectTestCommand(
+  workingDir: string,
+  readFile?: ReadFileFn,
+): Promise<DetectedTest | null> {
+  if (!readFile) return null;
+
+  const fileExists = async (path: string): Promise<string | null> => {
+    try {
+      return await readFile(path);
+    } catch {
+      return null;
+    }
+  };
+
+  // 1. package.json — Node.js projects
+  const packageJson = await fileExists(`${workingDir}/package.json`);
+  if (packageJson) {
+    try {
+      const pkg = JSON.parse(packageJson);
+      if (pkg.scripts?.test && pkg.scripts.test !== 'echo "Error: no test specified" && exit 1') {
+        const deps = { ...pkg.dependencies, ...pkg.devDependencies };
+        let framework = 'npm test';
+        if (deps?.vitest) framework = 'vitest';
+        else if (deps?.jest) framework = 'jest';
+        else if (deps?.mocha) framework = 'mocha';
+        return { command: 'npm test', framework, confidence: 'high' };
+      }
+    } catch { /* invalid JSON, skip */ }
+  }
+
+  // 2. Cargo.toml — Rust projects
+  const cargoToml = await fileExists(`${workingDir}/Cargo.toml`);
+  if (cargoToml) {
+    return { command: 'cargo test', framework: 'cargo', confidence: 'high' };
+  }
+
+  // 3. go.mod — Go projects
+  const goMod = await fileExists(`${workingDir}/go.mod`);
+  if (goMod) {
+    return { command: 'go test ./...', framework: 'go test', confidence: 'high' };
+  }
+
+  // 4. Python — pytest or unittest
+  const pytestIni = await fileExists(`${workingDir}/pytest.ini`);
+  if (pytestIni) {
+    return { command: 'pytest', framework: 'pytest', confidence: 'high' };
+  }
+  const pyprojectToml = await fileExists(`${workingDir}/pyproject.toml`);
+  if (pyprojectToml && pyprojectToml.includes('[tool.pytest')) {
+    return { command: 'pytest', framework: 'pytest', confidence: 'high' };
+  }
+
+  // 5. Makefile with test target
+  const makefile = await fileExists(`${workingDir}/Makefile`);
+  if (makefile && /^test\s*:/m.test(makefile)) {
+    return { command: 'make test', framework: 'make', confidence: 'medium' };
+  }
+
+  return null;
+}

package/src/council-engine/pipeline/types.ts

@@ -0,0 +1,401 @@
+/**
+ * Pipeline System: Type Definitions
+ * Councils as Workflow Steps — ordered stages with parallel steps
+ */
+
+// ============================================================================
+// Step & Pipeline Status
+// ============================================================================
+
+/** All pipeline step types. Gate, script, and condition are non-council types. */
+export const PIPELINE_STEP_TYPES = ['council', 'code_planning', 'analysis', 'agent', 'coding', 'review', 'enrich', 'gate', 'script', 'condition'] as const;
+export type PipelineStepType = (typeof PIPELINE_STEP_TYPES)[number];
+
+/** Council-based step types (excludes gate, script, condition). */
+export const COUNCIL_STEP_TYPES = ['council', 'code_planning', 'analysis', 'agent', 'coding', 'review', 'enrich'] as const;
+export type CouncilStepType = (typeof COUNCIL_STEP_TYPES)[number];
+
+/** Human-readable labels for each step type */
+export const STEP_TYPE_LABELS: Record<PipelineStepType, string> = {
+  council: 'Council',
+  code_planning: 'Code Planning',
+  analysis: 'Analysis',
+  agent: 'Agent',
+  coding: 'Coding',
+  review: 'Review',
+  enrich: 'Enrich',
+  gate: 'Gate',
+  script: 'Script',
+  condition: 'Condition',
+};
+
+/** What kind of data a step produces for downstream consumption */
+export type OutputType = 'string' | 'file' | 'directory' | 'json';
+
+export type PipelineStepStatus =
+  | 'pending'
+  | 'running'
+  | 'completed'
+  | 'failed'
+  | 'skipped'
+  | 'waiting_approval';
+
+export type PipelineStatus =
+  | 'draft'
+  | 'ready'
+  | 'running'
+  | 'completed'
+  | 'failed'
+  | 'paused';
+
+// ============================================================================
+// Step Artifacts (output produced by a completed step)
+// ============================================================================
+
+export interface StepArtifact {
+  stepId: string;
+  content: string;
+  artifactType: 'decision' | 'output' | 'llm_response' | 'approval';
+  metadata?: {
+    councilId?: string;
+    decisionId?: string;
+    outputId?: string;
+    model?: string;
+    tokensUsed?: number;
+    outputPath?: string;   // file path where output was saved (planning/coding steps)
+    outputType?: OutputType; // what kind of data this artifact represents
+    stepName?: string;     // human-readable name of the producing step
+    stepType?: string;     // PipelineStepType value
+  };
+  createdAt: string;
+}
+
+// ============================================================================
+// Memory Entry (persistent across scheduled runs)
+// ============================================================================
+
+export interface MemoryEntry {
+  runNumber: number;
+  runDate: string;
+  /** Captured artifacts from designated steps, keyed by sanitized step name */
+  captures: Record<string, string>;
+  /** True if this entry is a compressed summary of multiple older entries */
+  compressed?: boolean;
+}
+
+// ============================================================================
+// Output Isolation Config
+// ============================================================================
+
+export interface PipelineOutputConfig {
+  /** Enable structured output directories (default: true for new pipelines) */
+  enabled: boolean;
+  /** Max runs to keep on disk before pruning oldest (default: 50, 0 = unlimited) */
+  maxRetainedRuns?: number;
+  /** What to save per step */
+  stepOutput: 'artifact_only' | 'artifact_and_deliberation';
+}
+
+/** Run manifest written to _manifest.json */
+export interface RunManifest {
+  runNumber: number;
+  pipelineId: string;
+  pipelineName: string;
+  startedAt: string;
+  completedAt: string;
+  status: 'completed' | 'failed';
+  initialInput: string;
+  stageCount: number;
+  stepCount: number;
+  totalTokens: number;
+  totalDurationMs: number;
+  /** Whether memory was updated from this run */
+  memoryUpdated: boolean;
+}
+
+/** Step metadata written to _meta.json */
+export interface StepMeta {
+  stepId: string;
+  stepName: string;
+  stepType: string;
+  stageIndex: number;
+  stepIndex: number;
+  startedAt?: string;
+  completedAt?: string;
+  status: PipelineStepStatus;
+  outputType: OutputType;
+  councilId?: string;
+  model?: string;
+  tokensUsed?: number;
+  durationMs?: number;
+}
+
+// ============================================================================
+// Pipeline Persona (full-featured, matches council persona capabilities)
+// ============================================================================
+
+export interface PipelinePersona {
+  templateId?: string;
+  name: string;
+  role: 'manager' | 'consultant' | 'worker' | 'reviewer';
+  model: string;
+  provider: string;
+  avatar?: string;
+  color?: string;
+  systemPrompt?: string;
+  stance?: 'advocate' | 'critic' | 'neutral' | 'wildcard';
+  traits?: string[];
+  interactionStyle?: 'debate' | 'build' | 'question' | 'synthesize' | 'review';
+  domain?: string;
+  temperature?: number;
+  verbosity?: 'concise' | 'balanced' | 'thorough';
+  focusArea?: string;
+  startingStance?: string;
+  suppressPersona?: boolean;
+  /** Worker-only: save the worker's output to the working directory (default: true) */
+  saveOutput?: boolean;
+  /** MCP servers this persona can access (undefined = all servers) */
+  allowedServerIds?: string[];
+  /** Override default tool behavior: 'full' = enable all tools, 'none' = disable tools */
+  toolAccess?: 'full' | 'none';
+}
+
+// ============================================================================
+// Step Configs
+// ============================================================================
+
+export interface CouncilStepConfig {
+  type: CouncilStepType;
+  councilSetup: {
+    name: string;
+    personas: PipelinePersona[];
+    maxRounds?: number;
+    maxRevisions?: number;
+    expectedOutput?: string;
+    decisionCriteria?: string[];
+    workingDirectory?: string;
+    directoryConstrained?: boolean;
+    // Coding orchestrator config
+    testCommand?: string;
+    maxDebugCycles?: number;
+    maxReviewCycles?: number;
+    /** MCP servers this step can access (undefined = all servers) */
+    allowedServerIds?: string[];
+  };
+  /** Standing instructions — what this step should DO (supplemental to input context) */
+  task?: string;
+  /** Template that renders previous step outputs as input context */
+  inputTemplate: string;
+  /** What kind of data this step produces (default: 'string') */
+  outputType?: OutputType;
+  /** Include pipeline's initial input as additional context, regardless of stage */
+  includePipelineInput?: boolean;
+}
+
+export interface LlmStepConfig {
+  type: 'analysis' | 'agent';
+  model: string;
+  provider: string;
+  systemPrompt: string;
+  inputTemplate: string;
+  workingDirectory?: string;
+  directoryConstrained?: boolean;
+  /** MCP servers this step can access (undefined = all servers) */
+  allowedServerIds?: string[];
+}
+
+export interface GateStepConfig {
+  type: 'gate';
+  approvalPrompt: string;
+}
+
+export interface ScriptStepConfig {
+  type: 'script';
+  /** Shell command to execute. Previous step output available as $KONDI_INPUT env var. */
+  command: string;
+  /** Template that renders previous step outputs into the $KONDI_INPUT env var */
+  inputTemplate: string;
+  /** What kind of data this step produces (default: 'string') */
+  outputType?: OutputType;
+  /** Include pipeline's initial input as additional context, regardless of stage */
+  includePipelineInput?: boolean;
+}
+
+export type ConditionMode = 'contains' | 'regex' | 'equals';
+export type ConditionAction = 'continue' | 'skip_next_stage' | 'stop';
+
+export interface ConditionStepConfig {
+  type: 'condition';
+  /** The expression to match against (string literal, regex pattern, or exact match) */
+  expression: string;
+  /** How to evaluate the expression against input */
+  mode: ConditionMode;
+  /** Template that renders previous step outputs as input to evaluate */
+  inputTemplate: string;
+  /** Action when expression matches */
+  trueAction: ConditionAction;
+  /** Action when expression does not match */
+  falseAction: ConditionAction;
+  /** Include pipeline's initial input as additional context, regardless of stage */
+  includePipelineInput?: boolean;
+}
+
+export type StepConfig = CouncilStepConfig | LlmStepConfig | GateStepConfig | ScriptStepConfig | ConditionStepConfig;
+
+/** Helper: is this a council-based step type? Excludes gate, script, and condition. */
+export function isCouncilType(type: PipelineStepType): boolean {
+  return type !== 'gate' && type !== 'script' && type !== 'condition';
+}
+
+/** Helper: is this a lightweight council (single-agent, 0-round)? */
+export function isLightweightCouncilType(type: PipelineStepType): boolean {
+  return type === 'analysis' || type === 'agent';
+}
+
+/**
+ * @deprecated Use isLightweightCouncilType instead.
+ * Kept for backwards compatibility with existing code.
+ */
+export function isLlmType(type: PipelineStepType): boolean {
+  return type === 'analysis' || type === 'agent';
+}
+
+/**
+ * Migrate legacy LlmStepConfig to CouncilStepConfig.
+ * Old pipelines may have { type: 'analysis'|'agent', model, provider, systemPrompt, ... }
+ * without a councilSetup. This creates one from the flat fields.
+ */
+export function migrateLlmConfig(config: LlmStepConfig): CouncilStepConfig {
+  const isAnalysis = config.type === 'analysis';
+  return {
+    type: config.type,
+    councilSetup: {
+      name: isAnalysis ? 'Analysis' : 'Agent',
+      personas: [{
+        name: isAnalysis ? 'Analyst' : 'Executor',
+        role: isAnalysis ? 'manager' : 'worker',
+        model: config.model,
+        provider: config.provider,
+        systemPrompt: config.systemPrompt,
+        suppressPersona: true,
+      }],
+      maxRounds: 0,
+      maxRevisions: 0,
+      workingDirectory: config.workingDirectory,
+      directoryConstrained: config.directoryConstrained,
+      allowedServerIds: config.allowedServerIds,
+    },
+    inputTemplate: config.inputTemplate,
+    outputType: 'string',
+  };
+}
+
+// ============================================================================
+// Pipeline Step
+// ============================================================================
+
+export interface PipelineStep {
+  id: string;
+  name: string;
+  description?: string;
+  config: StepConfig;
+  status: PipelineStepStatus;
+  artifact?: StepArtifact;
+  error?: string;
+  startedAt?: string;
+  completedAt?: string;
+}
+
+// ============================================================================
+// Pipeline Stage
+// ============================================================================
+
+export interface PipelineStage {
+  id: string;
+  name: string;
+  steps: PipelineStep[];
+  /** How steps in this stage are executed (default: 'sequential') */
+  executionMode?: 'sequential' | 'parallel';
+}
+
+// ============================================================================
+// Pipeline
+// ============================================================================
+
+// ============================================================================
+// Schedule Config
+// ============================================================================
+
+export interface PipelineSchedule {
+  enabled: boolean;
+  /** Time of day in HH:MM (24-hour) format */
+  time: string;
+  /** Recurrence mode */
+  mode: 'once' | 'daily' | 'weekly';
+  /** For 'once' mode: ISO date string (YYYY-MM-DD) */
+  date?: string;
+  /** For 'weekly' mode: 0 = Sunday, 1 = Monday, ... 6 = Saturday */
+  dayOfWeek?: number;
+  /** ISO timestamp of the last scheduled run (to avoid double-fires) */
+  lastRunAt?: string;
+  /** Maintain memory across scheduled runs */
+  maintainMemory?: boolean;
+  /** Max detailed entries before older ones get compressed (default: 30) */
+  maxDetailedEntries?: number;
+  /** Which steps' artifacts to capture into memory (step IDs).
+   *  If empty/undefined, captures only the last completed step. */
+  captureStepIds?: string[];
+}
+
+export interface Pipeline {
+  id: string;
+  name: string;
+  description?: string;
+  initialInput: string;
+  stages: PipelineStage[];
+  settings: {
+    workingDirectory?: string;
+    directoryConstrained?: boolean;
+    failurePolicy: 'stop' | 'skip_step';
+    schedule?: PipelineSchedule;
+    outputConfig?: PipelineOutputConfig;
+  };
+  status: PipelineStatus;
+  currentStageIndex: number;
+  createdAt: string;
+  updatedAt: string;
+  /** Where this pipeline was executed — 'cli' for CLI-imported sessions */
+  source?: 'cli' | 'gui';
+}
+
+// ============================================================================
+// CLI ↔ GUI Session Export/Import
+// ============================================================================
+
+export interface KondiSessionCouncilData {
+  ledgerIndex: any;
+  ledgerChunks: Record<number, any[]>;
+  context: any | null;
+  contextHistory: any[];
+  contextPatches: any[];
+  decision: any | null;
+  plan: any | null;
+  directive: any | null;
+  outputs: any[];
+}
+
+export interface KondiSession {
+  version: 1;
+  exportedAt: string;
+  source: 'cli';
+  pipeline: Pipeline;
+  councils: any[];
+  councilData: Record<string, KondiSessionCouncilData>;
+  execution: {
+    status: 'completed' | 'failed';
+    startedAt: string;
+    completedAt: string;
+    durationMs: number;
+    workingDirectory: string;
+  };
+}

package/src/council-engine/services/deliberationSummary.ts

@@ -0,0 +1,114 @@
+/**
+ * Deliberation Summary Builder (Tauri-free)
+ * Extracted from deliberationSaveService.ts so it can be used by the pipeline
+ * executor without Tauri dependencies.
+ */
+
+import type { Council } from '../council/types';
+import { getAllEntries, buildMechanicalSummary } from '../council/ledger-store';
+import { getDecision, getLatestOutput } from '../council/context-store';
+
+// ============================================================================
+// Full Deliberation Builder
+// ============================================================================
+
+export function buildFullDeliberation(council: Council): string {
+  const entries = getAllEntries(council.id);
+  if (entries.length === 0) return '# Deliberation\n\nNo entries recorded.';
+
+  // Group entries by round
+  const rounds = new Map<number | 'none', typeof entries>();
+  for (const entry of entries) {
+    const key = entry.roundNumber ?? 'none';
+    if (!rounds.has(key)) rounds.set(key, []);
+    rounds.get(key)!.push(entry);
+  }
+
+  const getPersonaName = (personaId: string): string => {
+    const persona = council.personas.find((p) => p.id === personaId);
+    return persona?.name || personaId;
+  };
+
+  const getRoleName = (personaId: string): string => {
+    const assignment = council.deliberation?.roleAssignments?.find(
+      (r) => r.personaId === personaId
+    );
+    return assignment?.role || 'unknown';
+  };
+
+  let md = `# Deliberation: ${council.name}\n\n`;
+  md += `**Topic:** ${council.topic}\n`;
+  md += `**Date:** ${new Date().toISOString().split('T')[0]}\n`;
+  md += `**Total Entries:** ${entries.length}\n\n---\n\n`;
+
+  // Non-round entries first
+  const noRound = rounds.get('none');
+  if (noRound && noRound.length > 0) {
+    for (const entry of noRound) {
+      md += `### ${getPersonaName(entry.authorPersonaId)} (${getRoleName(entry.authorPersonaId)}) — ${entry.entryType}\n`;
+      md += `*${new Date(entry.timestamp).toLocaleString()}*\n\n`;
+      md += `${entry.content}\n\n---\n\n`;
+    }
+  }
+
+  // Round entries
+  const sortedRounds = Array.from(rounds.keys())
+    .filter((k) => k !== 'none')
+    .sort((a, b) => (a as number) - (b as number));
+
+  for (const round of sortedRounds) {
+    md += `## Round ${round}\n\n`;
+    const roundEntries = rounds.get(round)!;
+    for (const entry of roundEntries) {
+      md += `### ${getPersonaName(entry.authorPersonaId)} (${getRoleName(entry.authorPersonaId)}) — ${entry.entryType}\n`;
+      md += `*${new Date(entry.timestamp).toLocaleString()}*\n\n`;
+      md += `${entry.content}\n\n---\n\n`;
+    }
+  }
+
+  return md;
+}
+
+// ============================================================================
+// Abbreviated Summary Builder
+// ============================================================================
+
+export function buildAbbreviatedSummary(council: Council): string {
+  const entries = getAllEntries(council.id);
+  if (entries.length === 0) return 'No deliberation entries.';
+
+  let summary = `=== Deliberation Summary: ${council.name} ===\n\n`;
+
+  // Consultant highlights (mechanical summary)
+  const mechanicalSummary = buildMechanicalSummary(entries);
+  if (mechanicalSummary) {
+    // Replace persona IDs with names in the mechanical summary
+    let namedSummary = mechanicalSummary;
+    for (const p of council.personas) {
+      namedSummary = namedSummary.replace(new RegExp(p.id, 'g'), p.name);
+    }
+    summary += `--- Consultant Highlights ---\n${namedSummary}\n\n`;
+  }
+
+  // Decision
+  const decision = getDecision(council.id);
+  if (decision) {
+    summary += `--- Decision ---\n${decision.content}\n`;
+    if (decision.acceptanceCriteria) {
+      summary += `\nAcceptance Criteria: ${decision.acceptanceCriteria}\n`;
+    }
+    summary += '\n';
+  }
+
+  // Output (truncated)
+  const output = getLatestOutput(council.id);
+  if (output) {
+    const maxLen = 2000;
+    const truncated = output.content.length > maxLen
+      ? output.content.slice(0, maxLen) + '\n\n[... truncated ...]'
+      : output.content;
+    summary += `--- Output (v${output.version}) ---\n${truncated}\n`;
+  }
+
+  return summary;
+}

package/src/council-engine/tsconfig.json

@@ -0,0 +1,16 @@
+{
+  "compilerOptions": {
+    "target": "ES2022",
+    "module": "ESNext",
+    "moduleResolution": "bundler",
+    "allowImportingTsExtensions": true,
+    "noEmit": true,
+    "strict": true,
+    "skipLibCheck": true,
+    "esModuleInterop": true,
+    "resolveJsonModule": true,
+    "isolatedModules": true,
+    "lib": ["ES2022"]
+  },
+  "include": ["."]
+}