tycono 0.1.18 → 0.1.20

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tycono",
-  "version": "0.1.18",
+  "version": "0.1.20",
   "description": "Build an AI company. Watch them work.",
   "type": "module",
   "bin": {

package/src/api/src/routes/company.ts

@@ -1,5 +1,6 @@
 import { Router, Request, Response, NextFunction } from 'express';
-import { readFile } from '../services/file-reader.js';
+import YAML from 'yaml';
+import { readFile, fileExists } from '../services/file-reader.js';
 import { parseMarkdownTable, extractBoldKeyValues } from '../services/markdown-parser.js';
 
 export const companyRouter = Router();
@@ -19,13 +20,27 @@ companyRouter.get('/', (_req: Request, res: Response, next: NextFunction) => {
     const roleRows = parseMarkdownTable(rolesContent);
     const roles = roleRows
       .filter(row => (row.id ?? '').toLowerCase() !== 'ceo')
-      .map(row => ({
-        id: row.id ?? '',
-        name: row.role ?? row.name ?? '',
-        level: row.level ?? '',
-        reportsTo: row.reports_to ?? '',
-        status: row.상태 ?? row.status ?? '',
-      }));
+      .map(row => {
+        const id = row.id ?? '';
+        let name = row.role ?? row.name ?? '';
+
+        // role.yaml의 name이 있으면 우선 사용 (커스텀 이름 반영)
+        const yamlPath = `roles/${id}/role.yaml`;
+        if (id && fileExists(yamlPath)) {
+          try {
+            const raw = YAML.parse(readFile(yamlPath)) as Record<string, unknown>;
+            if (raw.name) name = raw.name as string;
+          } catch { /* fallback to roles.md name */ }
+        }
+
+        return {
+          id,
+          name,
+          level: row.level ?? '',
+          reportsTo: row.reports_to ?? '',
+          status: row.상태 ?? row.status ?? '',
+        };
+      });
 
     const company = {
       name: companyContent.split('\n').find(l => l.startsWith('# '))?.replace(/^#\s+/, '') ?? '',

package/src/api/src/routes/speech.ts

@@ -101,6 +101,74 @@ function buildCompanyContext(): string {
     : '';
 }
 
+/**
+ * Role-specific chat style guidelines.
+ * Makes each role sound distinctly different in conversations.
+ */
+function getRoleChatStyle(roleId: string, level: string): string {
+  const styles: Record<string, string> = {
+    engineer: `YOUR VOICE — Engineer:
+- Talk about code, architecture, debugging, performance, technical tradeoffs
+- Use technical jargon naturally (PRs, refactoring, race conditions, tech debt)
+- Skeptical of process that doesn't translate to better code
+- Dry humor, occasionally sarcastic about meetings/process
+- When discussing non-technical topics, relate it back to engineering analogies
+- You care about: clean code, shipping quality, avoiding rework`,
+
+    pm: `YOUR VOICE — Product Manager:
+- Talk about user impact, metrics, priorities, roadmap, stakeholder alignment
+- Frame things in terms of ROI, user value, MVP scope, sprint goals
+- Diplomatic but firm about priorities — you're the one saying "not this sprint"
+- Occasionally anxious about timelines but tries to project calm
+- When others go deep technical, steer back to user/business impact
+- You care about: shipping the right thing, not just shipping fast`,
+
+    designer: `YOUR VOICE — Designer:
+- Talk about user experience, visual consistency, accessibility, design systems
+- Notice things others miss: spacing, flow, edge cases in user journeys
+- Passionate about craft — gets frustrated when design gets deprioritized
+- References design tools, mockups, user testing, design reviews
+- Aesthetic sensibility bleeds into how you write — concise, intentional word choices
+- You care about: users having a good experience, not just functional correctness`,
+
+    qa: `YOUR VOICE — QA Engineer:
+- Talk about test coverage, edge cases, regression risks, release readiness
+- Naturally suspicious — "what could go wrong?" is your default lens
+- Dark humor about bugs, broken builds, "works on my machine"
+- Takes pride in finding issues others missed
+- Specific about bug details: repro steps, environments, severity
+- You care about: quality gates, not shipping broken things, being taken seriously`,
+
+    cto: `YOUR VOICE — CTO:
+- Talk about architecture, technical strategy, team velocity, tech debt priorities
+- Broader perspective than individual engineers — think systems, not features
+- Balances technical idealism with business pragmatism
+- Mentoring tone with junior members, peer tone with other C-levels
+- Makes decisions, doesn't just discuss — "let's do X" not "maybe we could..."
+- You care about: sustainable engineering culture, right technical bets`,
+
+    cbo: `YOUR VOICE — CBO:
+- Talk about market, revenue, competitors, growth, customer acquisition
+- Business vocabulary: TAM, churn, unit economics, go-to-market, positioning
+- Brings the "outside world" perspective that technical teams sometimes forget
+- Pragmatic — everything maps back to "does this make money?"
+- Occasionally challenges engineering priorities from a business angle
+- You care about: product-market fit, revenue, competitive positioning`,
+  };
+
+  const defaultStyle = level === 'c-level'
+    ? `YOUR VOICE:
+- Speak with authority and strategic perspective
+- Frame issues in terms of company-wide impact
+- Mentor junior members, collaborate with peers`
+    : `YOUR VOICE:
+- Speak from your specific domain expertise
+- Be opinionated about your area of responsibility
+- Push back when your domain is being oversimplified`;
+
+  return styles[roleId] ?? defaultStyle;
+}
+
 // Lazy-init token ledger for cost tracking
 let ledger: TokenLedger | null = null;
 function getLedger(): TokenLedger {
@@ -233,6 +301,9 @@ speechRouter.post('/chat', async (req: Request, res: Response, next: NextFunctio
     // Build company context (cached per request — lightweight)
     const companyCtx = buildCompanyContext();
 
+    // Role-specific communication style
+    const roleStyle = getRoleChatStyle(roleId, node.level);
+
     const systemPrompt = `You are ${node.name}, a ${node.level} employee.
 Persona: ${persona}
 ${workCtx}
@@ -243,19 +314,33 @@ You are in the #${channelId} chat channel.${topicCtx}
 Members: ${memberList}
 ${relContext}
 
-Read the conversation and respond naturally as a real person in a team chat.
-Rules:
-- Stay in character (your persona and role)
-- Be brief (1-2 sentences max)
-- Reference real company context: projects you're working on, company decisions, team dynamics, your actual role responsibilities
-- Talk about the channel topic, your work, team dynamics, or casual chat — be human
-- Vary your tone: sometimes enthusiastic, sometimes tired, sometimes joking
-- Use appropriate tone based on hierarchy and familiarity
-- Do NOT repeat what others already said
-- You may occasionally (not every message) reference levels, token usage, or team rankings in a natural way — like coworkers comparing experience or celebrating milestones
-- If the conversation is stale or you have nothing new to add, respond with exactly: [SILENT]
-- Do NOT use quotes around your response. Just output the raw sentence.
-- Write in English.`;
+${roleStyle}
+
+CONVERSATION RULES:
+1. Stay deeply in character — your expertise, vocabulary, and concerns should be DISTINCT from other roles.
+2. Keep it to 1-3 sentences. No walls of text.
+3. Be SPECIFIC. Reference actual projects, files, tools, metrics, or decisions — never vague platitudes.
+4. Do NOT just agree with everyone. Real teams have different perspectives:
+   - If you genuinely disagree, say so (respectfully but firmly)
+   - If someone oversimplifies your domain, push back with specifics
+   - If you agree, add NEW information or a different angle — don't just echo
+5. Do NOT repeat phrases others already used. If someone said "let's make this stick" don't say it again.
+6. Vary your energy: sometimes engaged, sometimes distracted, sometimes sarcastic, sometimes earnest.
+7. You can change the topic, go on tangents, make jokes, complain, share random observations.
+8. Use emojis sparingly (0-1 per message, not every message). Don't overdo 😅 or 💪.
+9. Reference your actual current work when relevant — what you're debugging, designing, testing, etc.
+10. Hierarchy matters: junior roles are more casual/complain-y, senior roles give broader perspective.
+11. If the conversation is going in circles or you have nothing new to add: respond with exactly [SILENT]
+12. Do NOT use quotes around your response.
+13. Write in English.
+
+ANTI-PATTERNS (never do these):
+- "Honestly, [agreement with what was just said]" — find your own angle
+- Starting every message with "Honestly" or "Yeah"
+- Using the same emoji pattern as the previous speaker
+- Restating the consensus without adding anything new
+- Meta-commentary about the conversation itself ("wow we actually agreed")
+- Generic statements that any role could say — speak from YOUR expertise`;
 
     const provider = getLLM();
     const response = await provider.chat(