আদর
Adar ARCL — Product Demo
Starting...
Intro · 00 — The Origin
What is Adar?
An agentic AI platform built with care, precision, and purpose — for the communities that need it most.
আদর
Bengali · noun · /aː.d̪or/
Affection, care, and thoughtful attention — the quality of tending to someone with warmth and precision.
ADAR stands for
Agentic
Data
Access &
Reasoning
🏏
Born from community passion — built to support recreational cricketers in Washington & the Greater Seattle area with meaningful data insights.
🌐
General-purpose platform — ARCL cricket is the reference implementation. The same platform powers any domain, any league, any community.
Intro · 01 — The Name
ADAR — Agentic Data Access and Reasoning
Every word in the name is intentional. Together they describe exactly what this system does — and why it is different from a simple chatbot.
A
Agentic
Autonomous agents that decide which tool to use, when to use it, and how to chain multiple calls together
D
Data
Live data fetched from arcl.org on demand — player stats, scorecards, schedules, rules — not a static database
A
Access
Multi-source access — arcl.org pages, Firestore vector search, Cloud SQL sessions — unified behind one interface
R
Reasoning
Gemini reasons over retrieved data — computes strength scores, aggregates seasons, interprets dismissals — not just retrieval
Applied to cricket: Adar is not a search engine
✗
Search engine: "Jiban Adhikary runs" → shows a TeamStats page
✓
Adar: fetches all seasons, aggregates, ranks, explains — then adds bowling context and dismissal breakdown
✗
Rule lookup: returns entire women's league rulebook for a men's question
✓
Adar: detects "men's league" intent, filters to correct rulebook, cites the exact section
Intro · 02 — Multi-Agent Architecture
Five agents, one orchestrator
Instead of one large model trying to do everything, Adar uses five specialised agents — each an expert in one domain. The orchestrator routes every question to the right expert.
User question
"How was Jiban dismissed in Spring 2026?"
"What is the wide rule for men's league?"
"Show batting stats as a graph"
🧭
Orchestrator
Routes to best agent
gemini-2.5-flash
gemini-2.5-flash
↕
🏏
Team Agent
Players · Scorecards · Dismissals · Strength
👤
Player Agent
Career stats · Season breakdown · Rankings
📖
Rules Agent
Men's · Women's · Umpiring guidelines
📡
Live Agent
Standings · Schedule · Recent results
⚖️
Judge Agent
Scores every response — accuracy, format, relevance
Intro · 03 — How Agents Communicate
Tool calling — agents reaching into the world
Agents do not just generate text. They call tools — functions that fetch real data. Each tool call is a deliberate action: fetch this page, search this collection, aggregate these seasons.
Example: "How was Jiban dismissed in Spring 2026?"
1
Orchestrator routes → Team Agent
"dismissal" keyword → team_agent handles scorecards
2
Tool call:
get_team_schedule("Agomoni Tigers", "Spring 2026")Fetches LeagueSchedule.aspx → returns 8 match IDs
3
Tool call × 8:
get_match_scorecard(match_id)Fetches Matchscorecard.aspx for each match → parses How_out column
4
Reasoning: aggregate all dismissals → generate response
Caught 7, Bowled 5, Run Out 3, Not Out 4 — with top dismissers ranked
Team Agent Tools
get_team_players_live
get_team_career_stats
get_match_scorecard
get_player_dismissals
get_team_schedule
get_teams_in_division
get_team_career_stats
get_match_scorecard
get_player_dismissals
get_team_schedule
get_teams_in_division
Rules Agent Tools
vector_search_rules
get_rule_section
get_faq_answer
Firestore vector search
with league filtering
get_rule_section
get_faq_answer
Firestore vector search
with league filtering
Live Agent Tools
get_standings
get_schedule
get_recent_results
get_announcements
Always live from arcl.org
get_schedule
get_recent_results
get_announcements
Always live from arcl.org
Intro · 04 — Short & Long Term Memory
Two types of memory working together
Adar uses both short-term conversational memory and long-term structured memory — each stored differently, each serving a different purpose.
💬
Short-term memory
Session context · Cloud SQL (Postgres)
Stores the current conversation thread — every message in the session. Adar remembers what was said earlier in the chat, so follow-up questions work naturally.
Example conversation:
You: Show Agomoni Tigers batting stats
Adar: [returns full table]
You: Now show the bowling
Adar: Knows "Agomoni Tigers" from context — no need to repeat
Storage: Cloud SQL sessions DB · Expires: 30 min auto-logout · Scope: per user per session
🧠
Long-term memory
Knowledge base · Firestore vector search
Stores indexed knowledge that persists across all sessions — ARCL rules, player career overviews, team standings. Searched semantically using vector embeddings (768 dimensions, Gemini embedding model).
arcl_rulesRules + FAQ chunks · vector indexed
arcl_playersCareer overview · vector indexed
arcl_teamsStandings · W/L/pts cached
arcl_evalsQuality scores · audit trail
Third type — live ephemeral data: TeamStats.aspx, Matchscorecard.aspx, LeagueSchedule.aspx are never cached. They are fetched fresh on every query — because arcl.org updates match results in real time and player stats change after every game.
01 / 08 · Getting started
Simple team login
Teams self-register and are approved by the ARCL admin. One login gives access to the full Adar assistant — no separate apps needed.
আদর
Adar ARCL
Sign in to your team account
No account? Register your team
Powered by Adar · American Recreational Cricket League
How teams get access
1
Picks ARCL team from dropdown
Live list from arcl.org — no typos possible
2
Stripe checkout — 14-day free trial
Card entered once · auto-renews monthly
3
Auto-activated + welcome email sent
Zero admin approval · ready to chat immediately
✉
Forgot password?
Reset link emailed · 1-hour expiry · no admin needed
Admin can also create accounts directly — Complimentary · Basic · Standard · Unlimited · or Pending payment for self-pay
02 / 08 · Core experience
Ask anything, get instant answers
Natural language chat powered by Gemini AI. Adar understands cricket context — no dashboards, no SQL, no exports needed.
আদর
Adar ARCL
Agomoni Tigers
💳 Billing
আদর
Hi! I'm Adar, the ARCL cricket assistant.
I can help you with league rules, player statistics, team history, and schedules.
What would you like to know?
I can help you with league rules, player statistics, team history, and schedules.
What would you like to know?
Try asking:
What is the wide rule in ARCL?
Show Agomoni Tigers players in Spring 2026
Can a player play for two teams?
Who scored the most runs in Div H?
How does the points table work?
03 / 08 · Rules engine
Rules — tagged by league automatically
Three separate rulebooks indexed: Men's League, Mens League Rules, Women's League. Adar always returns the right one — tagged at ingestion time.
আদর
Adar ARCL
Agomoni Tigers
You
What is the wide rule in ARCL?
আদর
Men's League Wide Ball Rule (Section: Umpiring — General)
Here are the wide ball rules for both Men's and Women's leagues in ARCL, as per the Umpiring section:
• A white line or a marker (35 inches from the middle stump) on the offside will be drawn to help the umpires in giving wides
• It's a wide ball if the ball is on or away from the wide line
• It's a wide ball if the ball passes above the striker's head standing upright at the popping crease
• If the ball on the leg side outside the leg stump is a wide; if it's on the leg stump then it's not a wide
• A ball is a wide ball if it bounces above the head of the batsman
• If a wide is also dead, it will be declared a dead ball
• Maximum 4 extras allowed per over · Maximum 10 balls per over including wides and no-balls
Spring 2026: specific overrides apply for Women's League...
Here are the wide ball rules for both Men's and Women's leagues in ARCL, as per the Umpiring section:
• A white line or a marker (35 inches from the middle stump) on the offside will be drawn to help the umpires in giving wides
• It's a wide ball if the ball is on or away from the wide line
• It's a wide ball if the ball passes above the striker's head standing upright at the popping crease
• If the ball on the leg side outside the leg stump is a wide; if it's on the leg stump then it's not a wide
• A ball is a wide ball if it bounces above the head of the batsman
• If a wide is also dead, it will be declared a dead ball
• Maximum 4 extras allowed per over · Maximum 10 balls per over including wides and no-balls
Spring 2026: specific overrides apply for Women's League...
Accuracy 5/5Complete 5/5Relevance 4/5Format 4/5Overall 4.5/5
Rules.aspx
General ARCL Rules
Tagged [MEN'S LEAGUE] at ingestion
Mens_League_Rules.htm
Men's League Rules
Tagged [MEN'S LEAGUE] at ingestion
Womens_League_Rules.htm
Women's League Rules
Tagged [WOMEN'S LEAGUE] — never mixed
04 / 08 · Live player data
Career stats across all seasons
Adar aggregates batting and bowling stats across every season played. Real data — fetched live from arcl.org — no manual updates needed.
আদর
Adar ARCL
Agomoni Tigers
You
show my teams stats → agomoni tigers
আদর
Agomoni Tigers — Career Stats (All Seasons)
Batting Stats
Batting Stats
| Player | Role | Seasons | Innings | Total Runs | Strike Rate | Fours | Sixes |
|---|---|---|---|---|---|---|---|
| Jiban Adhikary | All-rounder | 4 | 19 | 436 | 76.1 | 18 | 6 |
| Anijit Roy | All-rounder | 4 | 13 | 303 | 100.7 | 19 | 7 |
| Madhusudan Banik | All-rounder | 4 | 20 | 174 | 78.4 | 13 | 4 |
| BRAJA KRISHNA DAS | Batter | 4 | 18 | 140 | 52.5 | 1 | 0 |
| Sujit Biswas | All-rounder | 4 | 15 | 146 | 69.5 | 5 | 0 |
| Swadesh Poddar | All-rounder | 4 | 7 | 59 | 73.8 | 3 | 1 |
Accuracy 5/5Complete 5/5Relevance 5/5Format 4/5Overall 4.85/5
05 / 08 · Analysis & charts
Team strength analysis + visual charts
Adar calculates a strength score from batting depth, bowling variety and win rate. Bar charts render automatically — batting and bowling toggle with one click.
আদর
Adar ARCL
Agomoni Tigers
আদর
Team Analysis — Agomoni Tigers
| Metric | Value |
|---|---|
| Strength Rating | Strong |
| Strength Score | 95.8 / 100 |
| Win Rate | 39.6% (9 wins, 23 losses) |
| Top Batter | Jiban Adhikary |
| Top Bowler | Madhusudan Banik |
| Best All-rounder | Sujit Biswas |
📊 Batting
📊 Bowling
Jiban Adhikary
436
Anijit Roy
303
Madhusudan Banik
174
BRAJA KRISHNA DAS
140
Sujit Biswas
146
Swadesh Poddar
59
Madhusudan Banik
20
Anijit Roy
19
Swadesh Poddar
16
Sujit Biswas
15
Jiban Adhikary
13
Accuracy 5/5Complete 5/5Relevance 5/5Format 5/5Overall 5.0/5
06 / 08 · Schedule & umpiring
Schedule, results and umpiring assignments
Live match schedule with results, grounds, umpiring assignments and upcoming fixtures — fetched directly from arcl.org.
আদর
Adar ARCL
Agomoni Tigers
You
Show Agomoni Tigers team schedule and umpiring assignments
আদর
Agomoni Tigers — Schedule (Spring 2026)
Played Matches
| Date | Ground | H/A | Opponent | Result | Umpire | Status |
|---|---|---|---|---|---|---|
| 04/05 | Marymoor Soccer #06 | Home | Seattle Shouryangas | Won | Royal Priesthood | Played |
| 04/19 | SealTac Park | Away | Vikings-Strikers | Loss | Redmond Challengers | Played |
| 05/03 | Park Field | Home | SEA-Chargers | Won | Royal Priesthood | Played |
Upcoming Matches
| Date | Ground | H/A | Opponent | Umpire | Status |
|---|---|---|---|---|---|
| 05/10 | Marymoor Soccer #06 | Home | Royal Priesthood | Phoenix | Upcoming |
| 05/17 | Marymoor Soccer #06 | Away | Everest Warriors B | Royal Priesthood | Upcoming |
Umpiring Assignments
| Date | Ground | Team 1 | Team 2 | Status |
|---|---|---|---|---|
| 04/13 | Park Field | Redmond Indians | Willow Warriors | Played |
| 05/03 | SealTac Park | Vikings-Strikers | Royal Priesthood | Upcoming |
Accuracy 5/5Complete 5/5Relevance 5/5Format 4/5Overall 4.85/5
07 / 08 · Billing & subscriptions
Stripe-powered auto-renewing subscriptions
Teams subscribe once — Stripe handles auto-renewal, invoices, and failed payment recovery. PCI DSS compliant. No card data touches Adar's servers.
Billing & Subscription
Current plan
None · No subscription
Today's usage0
Payments processed securely by Stripe · PCI DSS compliant
Choose a plan
Basic
50 messages/day · Stats · Rules
$10/mo
POPULAR
Standard
200 messages/day · Full stats · Polls · Scorecards
$12/mo
Unlimited
1000 messages/day · Everything
$30/mo
14-day free trial on all plans · Stripe auto-renewal · Invoices emailed automatically
08 / 08 · Quality assurance
Every answer is graded automatically
A second Gemini model scores every response. Scores appear as badges in the chat. Admin dashboard tracks quality trends and flags low-scoring answers for review.
Scoring dimensions
Accuracy (35%)
5/5
Completeness (25%)
4/5
Relevance (25%)
5/5
Format (15%)
4/5
Overall
4.65
Average across all responses: 4.65 / 5
🔍 Filter: Max score ≤ 2 · May 05–06 · 2 flagged responses found
Scores 1.85 & 1.45 → wrong agent routing identified → fixed → now 4.2+
Why it matters
✓
Catch hallucinations — surfaces responses where AI invented facts
📈
Track prompt quality — see which agents need improvement
💰
~$0.0004 per eval — negligible cost using Gemini Flash
🔬
Fully automatic — runs on every response, zero setup
Executive Summary
Adar by the numbers
What Adar delivers for ARCL teams — and what it costs to build and run.
5
AI Agents
Orchestrator + 4 specialists + 1 judge
4.65
Quality Score / 5
Auto-graded on every response
<2s
Response Time
Live data from arcl.org
$12
Per Team / Month
200 msg/day · 14-day trial
💬 Questions teams ask every week
"Who are our top batsmen this season?"
"How was Jiban dismissed in the last match?"
"What is the wide ball rule for men's league?"
"Show our schedule and umpiring assignments"
"What is our team strength across all seasons?"
"Show batting stats as a bar chart"
"Can a player play for two teams this season?"
"[Self-service] Register → subscribe → chat — zero admin"
💰 Cost per team / month
Cloud Run$2–5
Firestore$1–3
Gemini API (chat + eval)~$1.13
Firebase HostingFree
Total infra~$7–12
Margin at $12/mo$2–5 per team
⚙️ Built on
Google ADK
Gemini 2.5 Flash
FastAPI
Cloud Run
Firestore
Firebase
Stripe
Cloud SQL
React + MUI
আদর
🌐 arcl.tigers.agomoniai.com
Built for ARCL.
Ready for every league.
Multi-agent AI · Live arcl.org data · Stripe subscriptions · LLM-as-judge evaluation · Multi-tenant architecture
5
AI agents
10+
Seasons of data
4.65/5
Avg quality score
$12
Per team / month