TJTEK Experts — AI Expert Advisory Council

🤖 Platform Assistant Live Demo

Hi! I'm the TJTEK Experts assistant. Ask me anything about our platform, the 13 experts, or how our AI routing works!

The Expert Council

Click any expert to start a consultation.

Convene the Council

Submit a topic for multi-expert deliberation and synthesis.

Select Experts (Max 5, or leave empty for Auto-select)

Deep Think Web Search

Analytics Dashboard

Expert consultation activity and patterns.

Admin Dashboard

Manage Platform Access (0 Users)

Name	Email	Provider	Status	Created	Actions

Pre-Approved Google Login IDs

Add email addresses that will be automatically approved on first sign-in.

Token Usage & Cost Tracking

—

Total Queries

—

Total Tokens

—

Est. Cost (30d)

Cost by Model

Per-User Usage

User ID	Queries	Tokens	Est. Cost

Recent Activity

Time	User	Action	Expert	Tokens	Cost

LLM Model Configuration

Select which AI model powers expert consultations.

Current Model

Loading...

Select Model

Expert Persona Management

View, edit, add, or remove expert SKILL.md files. Changes are saved to Firestore and take effect immediately.

Loading experts...

System Architecture

How the experts.tjtek.ai platform works end-to-end.

🔐 Authentication Flow

👤 User

→

Google Sign-In

→

Firebase Auth

→

Cloud Run API

→

Firestore
users/{uid}

→

✅ Approved / ⏳ Pending

🧠 Expert Consultation Flow

Question

→

Web Search
(optional)

→

Expert Persona
SKILL.md Prompt

→

LLM Router
Task Classifier

→

AI Model
(see routing below)

→

Response

🔀 Multi-LLM Routing Strategy

The LLM Router intelligently assigns each task to the optimal model based on task type:

🟢 Gemini 2.5 (Flash/Pro)
Quick Consult · Council · Web Search
Routine consultations, council hearings, and web-grounded queries. Fast, reliable, cost-effective.

🟡 Hermes 3 (Llama 3.1 70B)
Deep Think Mode
Extended chain-of-thought reasoning for complex strategic analysis. Excels at multi-step logic and nuanced thinking.

🟣 Kimi / OpenClaw (128K)
Long Context Analysis
Handles ultra-long documents and context windows up to 128K tokens. Ideal for comprehensive data analysis.

LLM Router

→

quick_consult?
→ Gemini

|

deep_think?
→ Hermes

|

long_context?
→ Kimi

|

web_grounded?
→ Gemini

⚔️ Council Deliberation Flow

Topic

→

Select Experts
(3-5)

→

Parallel LLM Calls
Gemini · One per expert

→

Collect Statements

→

Chairman
Synthesis

→

Resolution

🏛️ Tech Stack

Frontend
Vanilla JS, ECharts, Firebase Auth SDK

Backend
Python FastAPI, Uvicorn

AI Models
Gemini 2.5 (Flash/Pro) · Hermes 3 (70B) · Kimi/OpenClaw (128K)

LLM Router
Task-based routing: auto-selects optimal model per query type

Expert Engine
13 SKILL.md persona files loaded as system prompts

Auth
Firebase Authentication (Google SSO)

Database
Cloud Firestore (users, usage, config, expert overrides)

Hosting
Google Cloud Run (auto-scaling, containerized)

Search
Web grounding via Google Search API

LLM Providers
Vertex AI · Together.xyz (Hermes) · Moonshot AI (Kimi)

DNS
Custom domain experts.tjtek.ai

Admin
Live LLM config + Expert MD editor via Firestore

📖 User Guide

Everything you need to know about using experts.tjtek.ai

🚀 Getting Started

Sign In — Click "Log In" and sign in with your Google account.
Account Approval — New accounts require admin approval. While waiting, you can explore with demo responses.
Navigate — Use the top navigation tabs to switch between Experts, Council, Dashboard, and Help.

🧠 Consulting an Expert

Browse Experts — Go to the Experts tab to see all 13 world-class AI expert advisors.
Click an Expert — Click any expert card to open a consultation.
Ask Your Question — Type your question (up to 5,000 characters).
Configure Options:
- Deep Think — Enables extended reasoning for complex questions (slower but more thorough).
- Web Search — Grounds the expert's analysis with current web data.
- Language — Choose Auto (detects from question), English, or 中文.
Review Response — Read the expert's analysis, thinking chain, and search context.

⚔️ Convening the Council

Go to Council — Click the Council tab.
Enter a Topic — Describe your strategic challenge or question.
Select Experts — Choose up to 5 experts, or leave empty for auto-selection (the system picks 3 based on your topic).
Convene — Click "Convene Council" and wait for deliberation.
Read Results — Each expert provides their perspective, followed by the Chairman's Resolution — a synthesized verdict with action items.

👥 The 13 Experts

❓ FAQ

What is Deep Think?

Deep Think enables extended chain-of-thought reasoning. The AI spends more time analyzing your question from multiple angles before responding. It's slower but produces more thorough, nuanced analysis — ideal for complex strategic decisions.

What is Web Search grounding?

When enabled, the system searches the web for current information relevant to your question and provides it as context to the expert. This ensures responses reflect the latest data, news, and market conditions.

What languages are supported?

The platform supports English and Chinese (中文). Select "Auto" to let the system detect the language from your question, or explicitly choose English or 中文 from the dropdown.

What's the difference between demo and real responses?

If your account is pending approval, you'll receive demo responses — pre-written samples that show the format and style of real analysis. Once approved, you'll receive live AI-generated responses powered by Google Gemini.

How does the Council work?

The Council convenes multiple expert perspectives on a single topic. Each expert provides their analysis independently (in parallel), then a Chairman synthesizes all viewpoints into a unified resolution with a verdict, key risks, and actionable recommendations.

Is my data private?

Your questions and responses are processed through Google Cloud infrastructure. We do not store your consultation history beyond temporary session data. Usage metrics (token counts) are tracked for billing purposes only.