StockMate AI: Privacy-First Natural Language Inventory Assistant

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

Small business side hustles — such as boutique apparel shops, artisanal candle makers, and independent coffee roasters — run on razor-thin margins and packed schedules. Most owners juggle inventory between handwritten notes, chaotic spreadsheets, or clunky enterprise ERPs designed for mega-warehouses.

When boutique owners turn to generic cloud AI tools for help, they face two critical dealbreakers:

1. Hallucination Risk: Cloud LLMs invent numbers or round off inventory counts, causing costly stockouts or ghost stock.
2. Privacy & Data Leakage: Uploading confidential supplier invoices, wholesale discount tiers, wholesale costs, and warehouse passcodes to proprietary third-party servers leaks proprietary business advantages.

StockMate AI is a privacy-first, open-source inventory assistant powered by Google Gemma 2 and MongoDB Atlas, built specifically for side hustles and boutique retailers.

Key Capabilities:

• 🗣️ Natural Language Stock Assistant: Shop owners speak or type casually (e.g., "Received 15 linen shirts from supplier" or "Sold 3 ceramic mugs in store").
• 🛡️ Zero-Hallucination Safe Mode: Gemma classifies intent and extracts quantities, but transactions are validated by a deterministic Python state engine and confirmed via a human-in-the-loop modal before committing to the ledger.
• 🔒 Zero-Leak Privacy Whitelist: Before anything reaches Gemma, sensitive data (supplier emails, purchase contract notes, wholesale costs, warehouse coordinates) is stripped. Only sanitized operational metrics (e.g. Item-Apparel-P-101) are processed.
• ⚠️ Smart Restocking with Uncertainty Modeling: Proactively alerts owners when stock breaches safety thresholds, providing explicit uncertainty levels (Low, Medium, High) based on sales velocity.
• 📊 Grounded Business Q&A: Answers questions like "What should I restock this week?" or "What is my total inventory valuation?" grounded in verified MongoDB aggregation numbers.
• 📜 Immutable Movement Ledger: Automatically logs all sales, restocks, and transit losses with permanent timestamps.

Who Would Pay For It?
Independent boutique retailers, craft creators on Etsy/Shopify, and specialty roasters. StockMate AI saves 5+ hours of tedious weekly stock reconciliation while ensuring trade secrets and customer margins never leak. (I will keep it available as open source (the hosted solution might go down if my free quota expires or any thing else happened, but code is free and will be available in github.) And local setup is fully working, so just run it :) )

Demo

Deployed:

stockmate.chiranjib.tech

The above website is live thanks to Render.

• 🌐 Live Web Application: (If db not loading wait for a min and refresh , it should fetch from atlas) https://stockmate.chiranjib.tech
• 📡 Interactive API Docs: https://api.stockmate.chiranjib.tech/docs

(Hosted on Render with custom domain and automatic Let's Encrypt SSL).

Code

chiranjib2001 / StockMate-AI

StockMate AI 📦🤖

Privacy-First, Open-Source AI Inventory Assistant for Small Businesses & Side Hustles
Built for DEV Hacktoberfest 2026 Open-Source AI Challenge: Week 2 (Side Hustle)

---

💡 The Core Idea & The "Side Hustle" Dilemma

Solo entrepreneurs, boutique retailers, Instagram sellers, and artisan makers are running real businesses on razor-thin margins. Most still manage thousands of dollars in stock using chaotic spreadsheets, paper notebooks, or mental memory.

The Pain Points:

1. Costly Stockouts: High-velocity best-sellers suddenly hit zero stock before anyone notices, causing missed sales and frustrated customers.
2. Tied-Up Working Capital: Over-ordering slow-moving inventory because reorder calculations and sales velocity math are tedious and neglected.
3. The AI Privacy Paradox: Small business owners want AI assistance, but commercial platforms demand direct database access or require uploading sensitive supplier emails, wholesale invoices, customer details, and warehouse access codes.

The Solution: StockMate AI

StockMate AI is an intelligent, privacy-aware inventory operations platform designed specifically…

View on GitHub

• Repository: github.com/chiranjib2001/StockMate-AI
• License: MIT Open Source License

How I Built It

StockMate AI is architected as a hybrid open-weight intelligence pipeline:

flowchart TD
    User(["Boutique Owner"]) <--> UI["React 18 + Vite UI (Glassmorphism & Recharts)"]

    subgraph Backend ["FastAPI Backend"]
        API["REST API Gateway"]
        Router["Products · Movements · Analytics · AI"]
        API <--> Router

        subgraph PrivacyLayer ["Privacy Guard Layer"]
            Sanitizer["PII Regex Redactor"]
            Whitelist["Field Minimizer & Token Masker"]
            AuditLog["AI Audit Trail Logger"]
            Sanitizer --> Whitelist
        end

        subgraph DeterministicEngine ["Deterministic Safe Engine"]
            Rules["Inventory Availability Rules"]
            Math["Sales Velocity & Reorder Buffers"]
            Atomic["Atomic DB Transaction Commit"]
            Rules --> Math --> Atomic
        end
    end

    subgraph OpenWeightAI ["Open-Weight AI Pipeline"]
        Local["Local Ollama (Gemma 2:2b)"]
        Cloud["Cloud OpenRouter & Groq (Gemma 4 & Gemma 2)"]
    end

    subgraph DataLayer ["Data Layer"]
        DB[("MongoDB Atlas Replica Set")]
    end

    UI <--> API
    Router --> Sanitizer
    Whitelist --> Local
    Whitelist --> Cloud
    Local --> AuditLog
    Cloud --> AuditLog
    Router <--> DeterministicEngine
    DeterministicEngine <--> DB

Technology Stack:
Open-Weight AI (Google Gemma 2 & Gemma 4):
Local Private Mode: Runs offline via Ollama (gemma2:2b) with zero cloud data transmission.
Cloud Production Fallback (only for render we can't use localhost so need cloud providers): Powered by OpenRouter (google/gemma-4-26b-a4b-it:free) and Groq (gemma2-9b-it) with automatic multi-key rotation and rate-limit recovery.
Backend Engine:
FastAPI (Python 3.11) with async routing, Pydantic schemas, and structured error handling.
Async Motor Driver connecting to MongoDB Atlas.
Deterministic regex and token-overlap product matching to prevent zero-inventory edge cases.
Frontend Experience:
React 18 + TypeScript + Vite styled with a custom dark-mode Obsidian glassmorphism design system.
Recharts for real-time inventory category valuations and KPI distribution.
Lucide Icons for responsive micro-interactions.
Cloud Infrastructure & Domain:

• Render Web Service: Hosts the FastAPI backend container.
• Render Static Site: Deploys the optimized Vite React production build.
• .tech Domain: Live at stockmate.chiranjib.tech with automated TLS/SSL.

Why Does Open Innovation Matter?

True Privacy for Sensitive Trade Secrets: Closed proprietary AI APIs often reserve rights to use prompt data for model training. For a small boutique, supplier names, bulk pricing discounts, and wholesale margins are core trade secrets. By running open-weight Gemma with a strict privacy whitelist layer, shop owners know their business data remains entirely in their control.

Zero Marginal Operating Cost: A side hustle cannot afford $0.03 per prompt when logging dozens of small updates every day. Running Gemma locally on Ollama costs $0.00 in API fees, keeping overhead virtually zero.

Freedom from Lock-in: If a cloud provider changes pricing or deprecates an API version, open weights guarantee that the business logic continues running unmodified on any laptop or container.

My Agent Session

For dev session not adding.

Troubleshooting of the proj using antigravity

Agent

Created At: 2026-10-18T18:42:26+05:30
Completed At: 2026-10-18T18:42:26+05:30
File Path: file:///C:[REDACTED]/Videos/HACKATHON%20PROJ%20BOXES/StockMate-AI/backend/app/ai/gemma_service.py
Total Lines: 349
Total Bytes: 15436
Showing lines 18 to 135
The following code has been modified to include a line number before every line, in the format: <line_number>: <original_line>. Please note that any changes targeting the original code should remove the line number, colon, and leading space.
18: # 1. Cloud Provider: Groq (Open-weights fast cloud inference)
19: groq_key = settings.GROQ_API_KEY or (settings.GEMMA_API_KEY if settings.GEMMA_API_KEY.startswith("gsk_") else "")
20: if groq_key:
21: for model_id in ["llama-3.1-8b-instant", "llama-3.3-70b-versatile", "gemma2-9b-it"]:
22: try:
23: headers = {
24: "Authorization": f"Bearer {groq_key}",
25: "Content-Type": "application/json"
26: }
27: body: Dict[str, Any] = {
28: "model": model_id,
29: "messages": [
30: {"role": "system", "content": "You are a precise inventory assistant. Return only valid JSON without markdown wrapping if JSON is requested."},
31: {"role": "user", "content": prompt}
32: ],
33: "temperature": 0.2

Show more

Agent

[AI Reasoning]: Groq is attempted first if its API key is available; failures are handled gracefully by moving to the next provider. A crucial check is needed for OpenRouter. If the OPENROUTER_API_KEY environment variable isn't set, then the code skips OpenRouter entirely. The next step involves a local setup.

Local execution fails due to the absence of the local component in the remote environment, triggering a deterministic fallback. This fallback's broken logic, hardcoding quantity and action, results in inaccurate product matching. The objective is to implement a robust, offline-capable fallback strat

Show more

Agent

Created At: 2026-10-18T18:43:36+05:30
Completed At: 2026-10-18T18:43:36+05:30
File Path: file:///C:[REDACTED]/Videos/HACKATHON%20PROJ%20BOXES/StockMate-AI/backend/app/ai/gemma_service.py
Total Lines: 349
Total Bytes: 15436
Showing lines 140 to 240
The following code has been modified to include a line number before every line, in the format: <line_number>: <original_line>. Please note that any changes targeting the original code should remove the line number, colon, and leading space.
140: raw_user_input: str,
141: current_products: List[Dict[str, Any]]
142: ) -> Dict[str, Any]:
143: """
144: Interprets natural language statements such as:
145: 'I received 15 blue linen shirts and sold 3 today.'
146: """
147: # Step 1: PII Sanitization on raw user input
148: clean_input, pii_detected = redact_pii(raw_user_input)
149:
150: # Step 2: Build safe product reference context
151: safe_products, ref_to_id, privacy_audit = PrivacyMasker.build_safe_context(
152: current_products,
153: use_pseudonyms=False
154: )
155:
156: catalog_summary = [
157: {"product_name": p["product_name"], "current_stock": p["quantity_on_hand"], "ref": p["product_ref"]}
158: for p in safe_products
159: ]
160:
161: system_prompt = f"""You are StockMate AI's natural language inventory parser.
162: Extract the user's intended inventory chang

Show more

Agent

[AI Reasoning]: A helper method is being created to extract JSON payloads from text, handling potential markdown code block formatting. It will first strip whitespace, and then search for markdown code block delimiters and attempt to extract the JSON content.

Initial attempts at direct substring matching within product names or SKUs will be performed. If a match isn't found, an approach using token overlap between input and product catalog entries, excluding common stop words, will be considered. This dual-pass approach aims to improve accuracy in identifying the intended product.

Action intent is determi

Show more

Agent

Created At: 2026-10-18T18:44:25+05:30
Completed At: 2026-10-18T18:44:25+05:30
File Path: file:///C:[REDACTED]/Videos/HACKATHON%20PROJ%20BOXES/StockMate-AI/backend/app/ai/gemma_service.py
Total Lines: 349
Total Bytes: 15436
Showing lines 135 to 240
The following code has been modified to include a line number before every line, in the format: <line_number>: <original_line>. Please note that any changes targeting the original code should remove the line number, colon, and leading space.
135: print(f"[GemmaService] Ollama local call error: {e}")
136: return "", latency, str(e)
137:
138: async def parse_natural_language_intent(
139: self,
140: raw_user_input: str,
141: current_products: List[Dict[str, Any]]
142: ) -> Dict[str, Any]:
143: """
144: Interprets natural language statements such as:
145: 'I received 15 blue linen shirts and sold 3 today.'
146: """
147: # Step 1: PII Sanitization on raw user input
148: clean_input, pii_detected = redact_pii(raw_user_input)
149:
150: # Step 2: Build safe product reference context
151: safe_products, ref_to_id, privacy_audit = PrivacyMasker.build_safe_context(
152: current_products,
153: use_pseudonyms=False
154: )
155:
156: catalog_summary = [
157: {"product_name": p["product_name"], "current_stock": p["quantity_on_hand"], "ref": p["product_ref"]}
158:

Show more

Agent

Created At: 2026-10-18T18:45:01+05:30
Completed At: 2026-10-18T18:45:01+05:30
File Path: file:///C:[REDACTED]/Videos/HACKATHON%20PROJ%20BOXES/StockMate-AI/backend/app/ai/gemma_service.py
Total Lines: 349
Total Bytes: 15436
Showing lines 1 to 15
The following code has been modified to include a line number before every line, in the format: <line_number>: <original_line>. Please note that any changes targeting the original code should remove the line number, colon, and leading space.
1: import json
2: import time
3: import httpx
4: from typing import List, Dict, Any, Optional, Tuple
5: from datetime import datetime
6: from app.core.config import settings
7: from app.privacy.masking import PrivacyMasker, redact_pii
8: from app.database import db
9:
10: class GemmaService:
11: def init(self):
12: self.ollama_url = f"{settings.OLLAMA_BASE_URL}/api/generate"
13: self.model_name = settings.GEMMA_MODEL
14:
15: async def _call_gemma(self, prompt: str, is_json: bool = True) -> Tuple[str, float, Optional[str]]:
The above content does NOT show the entire file contents. If you need to view any lines of the file which were not shown to complete your task, call this tool again to view those lines.

Show more

Agent

[AI Reasoning]: Adding a regular expression import appears necessary for subsequent text processing within the service. Subsequently, several internal methods will enhance text analysis and intent classification; these include functions for JSON extraction, product matching, and deterministic intent extraction. Careful implementation of these methods is required to ensure reliable operation.

A method for extracting JSON from text is defined, handling potential markdown formatting and focusing on the outermost JSON object for parsing. A subsequent product-matching function utilizes both direct string matchin

Show more

Agent

Created At: 2026-10-18T18:45:54+05:30
Completed At: 2026-10-18T18:45:56+05:30
The following changes were made by the replace_file_content tool to: [REDACTED]\Videos\HACKATHON PROJ BOXES\StockMate-AI\backend\app\ai\gemma_service.py. If relevant, proactively run terminal commands to execute this code for the USER. Don't ask for permission.
[diff_block_start]
@@ -1,3 +1,4 @@
+import re
import json
import time
import httpx
@@ -11,6 +11,113 @@
self.ollama_url = f"{settings.OLLAMA_BASE_URL}/api/generate"
self.model_name = settings.GEMMA_MODEL

• def _extract_json(self, text: str) -> Optional[Dict[str, Any]]:
• if not text:
• return None
• clean = text.strip()
• if "```" in clean:
• match = re.search(r'(?:json)?\s*(\{[\s\S]*?\})\s*', clean)
• if match:
• clean = match.group(1)
• obj_match = re.search(r'({[\s\S]*})', clean)
• if obj_match:
• clean = obj_match.group(1)
• try:
• return json.loads(clean)
• except Exception:
• return None +
• def _match_product(self, target_name: str, clean_input: str, current_products: List[Dict[str, Any]]) -> Optional[Dict[str, Any]]:
• if not current_products:
• return None +
• target = target_name.strip().lower()
• full_text = clean_input.strip().lower() +
• # Pass 1: Direct substring or SKU match
• if target:
• for p in cu

Show more

Agent

Created At: 2026-10-18T18:46:58+05:30
Completed At: 2026-10-18T18:46:58+05:30
File Path: file:///C:[REDACTED]/Videos/HACKATHON%20PROJ%20BOXES/StockMate-AI/backend/app/ai/gemma_service.py
Total Lines: 457
Total Bytes: 19714
Showing lines 285 to 355
The following code has been modified to include a line number before every line, in the format: <line_number>: <original_line>. Please note that any changes targeting the original code should remove the line number, colon, and leading space.
285: "action": "sale" | "restock" | "adjustment" | "loss" | "new_product",
286: "product_name": "string",
287: "quantity": integer,
288: "confidence": float (0.0 to 1.0),
289: "reasoning": "string"
290: }}"""
291:
292: raw_resp, latency, err = await self._call_gemma(system_prompt, is_json=True)
293:
294: # Fallback if Ollama error
295: parsed = {}
296: if not err and raw_resp:
297: try:
298: parsed = json.loads(raw_resp)
299: except Exception:
300: parsed = {"action": "unknown", "product_name": "", "quantity": 0, "confidence": 0.0, "reasoning": raw_resp}
301: else:
302: parsed = {
303: "action": "adjustment",
304: "product_name": clean_input,
305: "quantity": 1,
306: "confidence": 0.5,
307: "reasoning": f"Local rule fallback (Ollama message: {err})"
308: }
309:
310: # St

Show more

Agent

[AI Reasoning]: Initial attempt involved extracting structured data from the AI's response. The primary hypothesis was that the response would conform to a predefined JSON format, which then needed parsing. An error check was performed after the function call and before parsing the response. If the AI output was successfully parsed into a dictionary, the next step would be processing.

Action and quantity need to be extracted from the parsed data, handling potential missing or non-numeric quantity values by defaulting to zero, and attempting a recovery with regex if necessary. Product matching relies on an i

Show more

Agent

Created At: 2026-10-18T18:47:50+05:30
Completed At: 2026-10-18T18:47:53+05:30
The following changes were made by the replace_file_content tool to: [REDACTED]\Videos\HACKATHON PROJ BOXES\StockMate-AI\backend\app\ai\gemma_service.py. If relevant, proactively run terminal commands to execute this code for the USER. Don't ask for permission.
[diff_block_start]
@@ -291,52 +291,65 @@

     raw_resp, latency, err = await self._call_gemma(system_prompt, is_json=True)

• # Fallback if Ollama error
• parsed = {}
• if not err and raw_resp:
• try:
• parsed = json.loads(raw_resp)
• except Exception:
• parsed = {"action": "unknown", "product_name": "", "quantity": 0, "confidence": 0.0, "reasoning": raw_resp}
• else:
• parsed = {
• "action": "adjustment",
• "product_name": clean_input,
• "quantity": 1,
• "confidence": 0.5,
• "reasoning": f"Local rule fallback (Ollama message: {err})"

- }

• # Step 3: Match with real product in DB
• target_name = parsed.get("product_name", "").lower()
• matched_prod = None
• for p in current_products:
• p_name = p.get("name", "").lower()
• if target_name in p_name or p_name in target_name:
• matched_prod = p

- break

- # If no exact match, fallback to first product or none

Show more

Agent

Created At: 2026-10-18T18:48:48+05:30
Completed At: 2026-10-18T18:48:48+05:30
File Path: file:///C:[REDACTED]/Videos/HACKATHON%20PROJ%20BOXES/StockMate-AI/backend/app/api/routes/ai.py
Total Lines: 115
Total Bytes: 3852
Showing lines 80 to 115
The following code has been modified to include a line number before every line, in the format: <line_number>: <original_line>. Please note that any changes targeting the original code should remove the line number, colon, and leading space.
80: """
81: Demonstrates how raw inventory records are privacy-filtered
82: before any data is transmitted to Gemma.
83: """
84: raw_sample = {
85: "sku": "LNN-SHT-BLU",
86: "name": "Classic Linen Button-Down (Ocean Blue)",
87: "category": "Apparel",
88: "unit_cost": 18.50,
89: "selling_price": 48.00,
90: "quantity_on_hand": 14,
91: "minimum_stock": 10,
92: "supplier_name": "Seaside Fabrics Co.",
93: "supplier_email": "[REDACTED]",
94: "warehouse_location": "Building 4, Secure Vault Bay 12",
95: "notes": "Contract discount expires Nov 2026. Gate passcode: 4892."
96: }
97:
98: minimized = PrivacyMasker.minimize_product_for_ai(raw_sample, "P-101", units_sold_30d=24)
99:

100: return {
101: "raw_record_with_pii": raw_sample,
102: "sanitized_ai_payload": minimized,
103: "redacted_fields": [
104: "supplier_n

Show more

Agent

Created At: 2026-10-18T18:51:26+05:30
Completed At: 2026-10-18T18:51:28+05:30
The following changes were made by the replace_file_content tool to: [REDACTED]\Videos\HACKATHON PROJ BOXES\StockMate-AI\backend\app\api\routes\ai.py. If relevant, proactively run terminal commands to execute this code for the USER. Don't ask for permission.
[diff_block_start]
@@ -112,4 +112,60 @@
@router.get("/audit-logs", response_model=List[AIAuditLog])
async def get_audit_logs(limit: int = 25):
return await db.get_ai_audits(limit=limit)
+
[e-mail]("/diagnostics")
+async def ai_diagnostics():
+ import httpx
+ from app.core.config import settings
+
+ diag: Dict[str, Any] = {}
+
+ # 1. Groq Check
+ groq_key = settings.GROQ_API_KEY or (settings.GEMMA_API_KEY if settings.GEMMA_API_KEY.startswith("gsk_") else "")
+ if groq_key:
+ masked_groq = groq_key[:8] + "..." + groq_key[-4:] if len(groq_key) > 12 else "present"
+ diag["groq"] = {"key": masked_groq}
+ try:
+ headers = {"Authorization": f"Bearer {groq_key}", "Content-Type": "application/json"}
+ body = {"model": "llama-3.1-8b-instant", "messages": [{"role": "user", "content": "ping"}], "max_tokens": 5}
+ async with httpx.AsyncClient(timeout=8.0) as client:
+ r = await client.post("https://api.groq.com/openai/v1/chat/completions", headers=headers, json=body)
+ diag["groq"]["status"] = r.status_code
+ diag["groq"]["r

Show more

Agent

Created At: 2026-10-18T18:54:35+05:30
Completed At: 2026-10-18T18:54:37+05:30

            The command exited with code 0.
            Output:
            Backend loaded successfully!

Agent

Created At: 2026-10-18T18:54:56+05:30
Completed At: 2026-10-18T18:54:56+05:30
Created file file:///C:[REDACTED]/.gemini/antigravity-ide/brain/cc551c86-c432-44be-867c-b5e545626744/scratch/test_nl_full.py with requested content.
If relevant, proactively run terminal commands to execute this code for the USER. Don't ask for permission.

Agent

Created At: 2026-10-18T18:55:13+05:30
Tool is running as a background task with task id: cc551c86-c432-44be-867c-b5e545626744/task-635
Task Description: python "[REDACTED].gemini\antigravity-ide\brain\cc551c86-c432-44be-867c-b5e545626744\scratch\test_nl_full.py"
Task logs are available at: file:///C:[REDACTED]/.gemini/antigravity-ide/brain/cc551c86-c432-44be-867c-b5e545626744/.system_generated/tasks/task-635.log
YOU MUST TAKE ONE OF THE FOLLOWING TWO ACTIONS: A) either proceed to other relevant work (if any) or, B) simply update the user with a short message (that you have launched the command and will wait for it to finish) and end the turn.
DO NOTHING ELSE.

Show more

Agent

Created At: 2026-10-18T18:55:34+05:30
Completed At: 2026-10-18T18:55:34+05:30
File Path: file:///C:[REDACTED]/.gemini/antigravity-ide/brain/cc551c86-c432-44be-867c-b5e545626744/.system_generated/tasks/task-635.log
Total Lines: 23
Total Bytes: 1266
Showing lines 1 to 23
The following code has been modified to include a line number before every line, in the format: <line_number>: <original_line>. Please note that any changes targeting the original code should remove the line number, colon, and leading space.
1: [DB] MongoDB not reachable (No servers found yet, Timeout: 1.5s, Topology Description: <TopologyDescription id: 6ad4c8baaf91bb292934d0f8, topology_type: Unknown, servers: [<ServerDescription ('localhost', 27017) server_type: Unknown, rtt: None>]>). Using persistent local JSON engine.
2:
3: --- Testing Query: 'Received 15 linen shirts from supplier' ---
4: Matched: Classic Linen Button-Down (Ocean Blue)
5: Action: restock
6: Delta: 15
7: New Stock: 64
8: Explanation: The user's statement mentions receiving linen shirts, aligning with the 'Classic Linen Button-Down (Ocean Blue)' product description. This indicates a restock action.
9:
10: --- Testing Query: 'Sold 3 ceramic mugs in store' ---
11: Matched: Handmade Speckled Ceramic Mug
12: Action: sale
13: Delta: -3
14: New Stock: 13
15: Explanation: The user explicitly states 'Sold 3 ceramic mugs in store', indicating a sale of the 'Handmade Speckled Ceramic Mug' with a quantity of 3 units.

Show more

Agent

[AI Reasoning]: The phrase "Received 15 linen shirts" strongly suggests an inventory restock event, and the product description matches linen shirts. Thus, the restock action, with a delta of 15, is confirmed.

Agent

Created At: 2026-10-18T18:55:54+05:30
Completed At: 2026-10-18T18:55:54+05:30

            The command exited with code 0.
            Output:
            On branch main

Your branch is up to date with 'origin/main'.

Changes not staged for commit:
(use "git add <file>..." to update what will be committed)
(use "git restore <file>..." to discard changes in working directory)
modified: backend/app/ai/gemma_service.py
modified: backend/app/api/routes/ai.py

no changes added to commit (use "git add" and/or "git commit -a")

Agent

Created At: 2026-10-18T18:56:46+05:30
Completed At: 2026-10-18T18:56:53+05:30

            The command exited with code 0.
            Output:
            warning: in the working copy of 'backend/app/ai/gemma_service.py', LF will be replaced by CRLF the next time Git touches it

warning: in the working copy of 'backend/app/api/routes/ai.py', LF will be replaced by CRLF the next time Git touches it
[main 0336c65] fix(ai): smart token product matching, robust markdown json extraction, and diagnostics endpoint
2 files changed, 220 insertions(+), 43 deletions(-)
To https://github.com/chiranjib2001/StockMate-AI.git
6d5f77e..0336c65 main -> main

Show more

Agent

Root Cause of the Bug

1. Brittle Fallback Logic: Previously, when an external LLM request had formatting wrappers or connection delays, the old fallback code was hardcoded with quantity: 1, action: "adjustment", and set product_name: "Received 15 linen shirts from supplier".
2. Substring Catalog Matching: The previous code checked if target_name in p_name. Because the full sentence "Received 15 linen shirts from supplier" was not an exact substring of "Classic Linen Button-Down (Ocean Blue)", it failed to match and fell back to "Unmatched / New Item" with +1 unit.
3. Markdown JSON Parsing: When LLMs wrapped their JSON in ```json ... ``` fences, json.loads() threw an exception, kicking the request into that broken fallback.

---

What Was Fixed (Commit 0336c65)

1. Smart Product Token Matcher (_match_product):
   • Computes semantic token overlap between query terms ("linen", "shirts") and catalog names, SKUs, and categories.
   • Accurately identifies "Classic Linen Button-Down (Ocean Blue)" even when phrased casually (e.g., "15 linen shirts").
2. Deterministic Safe-Mode Engine (_deterministic_intent_extract):
   • Quantity extraction: Uses regex r'\b(\d+)\b' to extract the actual quantity (15).
   • Action classification: Recognizes keywords (received/restocked ➔ restock, sold ➔ sale, broke ➔ loss).
   • Stock calculation: Accurately applies the delta (+15) to current sto

Show more

You

Help me submit it . I want to do it myshelf just guide me

CHALLANGE:
```
Skip to content
DEV Community
Search...
Powered by Algolia
Create Post

Challenges
→
Hacktoberfest Open-Source AI Challenge: Week 2
Hacktoberfest Open-Source AI Challenge: Week 2
Hacktoberfest Open-Source AI Challenge: Week 2
Signed Up
View Entries
Challenge 3 of 5 in our Hacktoberfest DEV Challenges. One prompt, a new theme every week.

Challenge Status:
Live
This year's Hacktoberfest is not about open source pull requests. There's no PR count to hit and no repos to hunt for. Instead, you build a brand-new project with open-source AI at its core, and write about it on DEV. Here's everything you need to know about Hacktoberfest 2026.

Week 2 of our five Hacktoberfest DEV Challenges starts today! Running through October 18, the Hacktoberfest Open-Source AI Challenge: Week 2 gives you a full week to build.

Missed Week 1: Touch Grass? No problem. Every challenge is a fresh start, so you're competing on the same footing as everyone else.

This week, we'll select one overall winner and a winner in each of our 17 partner categories. See all five challenges and current partner categories on the HF26 DEV Challenge Hub.

Free partner credits

Several partners are giving participants credits and promo codes, including Tinker, Render, Backboard, and a free .tech domain. Claim yours at hacktoberfest.com/my/promos.

Tinker by Thinking Machines
Render
Backboard
.tech Domains
Claim your credits →
Visit hacktoberfest.com
Key Dates
Contest start: October 12, 2026
Submissions due: October 18, 2026
Winners announced: Week of October 19
Badge Rewards
Hacktoberfest Challenge Completion Badge
Hacktoberfest Challenge Completion Badge
Hacktoberfest Challenge Winner Badge
Hacktoberfest Challenge Winner Badge
Find Out More
Ask questions and share your ideas on the Hacktoberfest Open-Source AI Challenge: Week 2 Launch Post.

View Launch Post
Challenge Prompt
Side

Show more

Agent

[AI Reasoning]: Understanding the user's intent: they desire guidance to independently submit "StockMate AI" to the "Side Hustle" Hacktoberfest challenge. This requires synthesizing submission guidelines with the provided project details.

The core task involves structuring the submission according to the specified template: cover image, title, and relevant tags, including the mandatory and additional challenge tags. Reviewing the provided materials should uncover all the necessary information for a complete submission.

Agent

Created At: 2026-10-18T19:10:53+05:30
Completed At: 2026-10-18T19:10:53+05:30
File Path: file:///C:[REDACTED]/.gemini/config/skills/devrelay-sessions/SKILL.md
Total Lines: 132
Total Bytes: 4791
Showing lines 1 to 132
The following code has been modified to include a line number before every line, in the format: <line_number>: <original_line>. Please note that any changes targeting the original code should remove the line number, colon, and leading space.
1: ---
2: name: devrelay-sessions
3: description: Save, curate, browse, and embed AI agent session transcripts on DEV (Forem) via DevRelay. Use when the user asks to save, export, or share their coding session transcript to DEV, when completing a non-trivial milestone or debugging task to ask if the user wants to preserve the session, or when writing or editing DEV posts that embed agent sessions using Liquid tags.
4: ---
5:
6: # DevRelay Agent Sessions
7:
8: DEV supports native Agent Sessions — rich interactive transcripts that capture prompts, thoughts, code changes, and tool calls. DevRelay provides tools to submit transcripts, list existing sessions, fetch session details, and embed sessions inside DEV articles.
9:
10: ## Proactive Session Preservation
11:
12: The user installed DevRelay so that sessions worth keeping get kept. You are the one who
13: notices the moment. When you reach the conclusion of a significant coding milestone,
14: architectural refactor, complex debugging session, or technical brea

Show more

Agent

[AI Reasoning]: Native agent session support appears relevant, alongside the user's expressed desire to shape the outcome in a specific way. The upload screen provides details about session files, formats (.jsonl, .json), and their typical location, which suggests a path for further investigation. This session information might allow for a focused approach to modifying the desired outcome.

Agent

Created At: 2026-10-18T19:11:12+05:30
Completed At: 2026-10-18T19:11:12+05:30

            The command exited with code 0.
            Output:
            True

Agent

[AI Reasoning]: The provided transcript file presents a significant security risk due to the potential inclusion of sensitive credentials like API keys and database passwords. This could lead to the exposure of confidential information if shared or uploaded. A method must be established to sanitize the file.

Sanitization of the session export file requires identifying and redacting sensitive data like API keys and database credentials. The goal is to produce a clean, formatted JSON file suitable for sharing, with placeholders replacing confidential information. This involves both data redaction and schema a

Show more

Agent

Created At: 2026-10-18T19:11:40+05:30
Completed At: 2026-10-18T19:11:40+05:30
Created file file:///C:[REDACTED]/.gemini/antigravity-ide/brain/cc551c86-c432-44be-867c-b5e545626744/scratch/create_clean_session.py with requested content.
If relevant, proactively run terminal commands to execute this code for the USER. Don't ask for permission.

Agent

Created At: 2026-10-18T19:11:56+05:30
Completed At: 2026-10-18T19:11:57+05:30

            The command exited with code 0.
            Output:
            Reading transcript from: [REDACTED]\.gemini\antigravity-ide\brain\cc551c86-c432-44be-867c-b5e545626744\.system_generated\logs\transcript.jsonl

Total curated messages: 0
Sanitized session file written to: [REDACTED]\Videos\HACKATHON PROJ BOXES\StockMate-AI\stockmate_agent_session_clean.json
File size: 470 bytes

Agent

Created At: 2026-10-18T19:12:11+05:30
Completed At: 2026-10-18T19:12:11+05:30
File Path: file:///C:[REDACTED]/.gemini/antigravity-ide/brain/cc551c86-c432-44be-867c-b5e545626744/.system_generated/logs/transcript.jsonl
Total Lines: 639
Total Bytes: 856097
Showing lines 1 to 2
The following code has been modified to include a line number before every line, in the format: <line_number>: <original_line>. Please note that any changes targeting the original code should remove the line number, colon, and leading space.
1: {"step_index":0,"source":"USER_EXPLICIT","type":"USER_INPUT","status":"DONE","created_at":"2026-10-13T14:23:50Z","content":"<USER_REQUEST>\nread baseidea.md\n</USER_REQUEST>\n<ADDITIONAL_METADATA>\nThe current local time is: 2026-10-13T19:53:50+05:30.\n\nThe user's current state is as follows:\nActive Document: c:\Users\<USERNAME MASK>\Videos\HACKATHON PROJ BOXES\StockMate-AI\baseidea.md (LANGUAGE_MARKDOWN)\nCursor is on line: 1\nOther open documents:\n- c:\Users\<USERNAME MASK>\Videos\HACKATHON PROJ BOXES\StockMate-AI\baseidea.md (LANGUAGE_MARKDOWN)\n</ADDITIONAL_METADATA>\n<USER_SETTINGS_CHANGE>\nThe user changed setting Model Selection from None to Gemini 3.8 Flash (High). No need to comment on this change if the user doesn't ask about it. If reporting what model you are, please use a human readable name instead of the exact string.\n</USER_SETTINGS_CHANGE>"}
2: {"step_index":1,"source":"SYSTEM","type":"CONVERSATION_HISTORY","status":"DONE","created_at":"

Show more

Agent

Created At: 2026-10-18T19:12:31+05:30
Completed At: 2026-10-18T19:12:31+05:30
File Path: file:///C:[REDACTED]/.gemini/antigravity-ide/brain/cc551c86-c432-44be-867c-b5e545626744/.system_generated/logs/transcript.jsonl
Total Lines: 641
Total Bytes: 858465
Showing lines 3 to 8
The following code has been modified to include a line number before every line, in the format: <line_number>: <original_line>. Please note that any changes targeting the original code should remove the line number, colon, and leading space.
3: {"step_index":2,"source":"SYSTEM","type":"KNOWLEDGE_ARTIFACTS","status":"DONE","created_at":"2026-10-13T14:23:50Z"}
4: {"step_index":3,"source":"MODEL","type":"PLANNER_RESPONSE","status":"DONE","created_at":"2026-10-13T14:23:50Z","thinking":"Checking if relevant knowledge items exist to inform this task before proceeding to read the file. Examining the available knowledge to see if any pre-existing information applies to the user's request. The user's intent is to examine baseidea.md.\n\n\n","tool_calls":[{"name":"view_file","args":{"AbsolutePath":"\"C:[REDACTED]/Videos/HACKATHON PROJ BOXES/StockMate-AI/baseidea.md\"","toolAction":"\"Reading baseidea.md\"","toolSummary":"\"Read baseidea.md\""}}]}
5: {"step_index":4,"source":"MODEL","type":"VIEW_FILE","status":"DONE","created_at":"2026-10-13T14:23:55Z","content":"Created At: 2026-10-13T19:53:55+05:30\nCompleted At: 2026-10-13T19:53:55+05:30\nFile Path: `file:///C:[REDACTED]/Videos/HACKATHON%20P

Show more

Agent

Created At: 2026-10-18T19:13:13+05:30
Completed At: 2026-10-18T19:13:13+05:30
Created file file:///C:[REDACTED]/.gemini/antigravity-ide/brain/cc551c86-c432-44be-867c-b5e545626744/scratch/build_clean_session.py with requested content.
If relevant, proactively run terminal commands to execute this code for the USER. Don't ask for permission.

Agent

Created At: 2026-10-18T19:13:29+05:30
Completed At: 2026-10-18T19:13:30+05:30

            The command exited with code 0.
            Output:
            Extracted 0 messages.

Saved sanitized file (0 messages) to [REDACTED]\Videos\HACKATHON PROJ BOXES\StockMate-AI\stockmate_agent_session_clean.json

Agent

Created At: 2026-10-18T19:13:53+05:30
Completed At: 2026-10-18T19:13:53+05:30

            The command exited with code 0.
            Output:
            Line 1 length: 875

Keys: dict_keys(['step_index', 'source', 'type', 'status', 'created_at', 'content'])
type: USER_INPUT
source: USER_EXPLICIT

Agent

Created At: 2026-10-18T19:14:11+05:30
Completed At: 2026-10-18T19:14:11+05:30

            The command exited with code 0.
            Output:
            raw content len: 722

match:
read baseidea.md

Agent

Created At: 2026-10-18T19:14:27+05:30
Completed At: 2026-10-18T19:14:28+05:30

            The command exited with code 0.
            Output:
            Total lines in file: 650

Agent

Created At: 2026-10-18T19:14:47+05:30
Completed At: 2026-10-18T19:14:47+05:30
Created file file:///C:[REDACTED]/.gemini/antigravity-ide/brain/cc551c86-c432-44be-867c-b5e545626744/scratch/build_clean_session.py with requested content.
If relevant, proactively run terminal commands to execute this code for the USER. Don't ask for permission.

Agent

Created At: 2026-10-18T19:15:07+05:30
Completed At: 2026-10-18T19:15:08+05:30

            The command exited with code 0.
            Output:
            Error on line 0: bad escape \U at position 2

Error on line 3: bad escape \U at position 2
Error on line 4: bad escape \U at position 2
Extracted 0 messages successfully.
Saved 0 curated messages to [REDACTED]\Videos\HACKATHON PROJ BOXES\StockMate-AI\stockmate_agent_session_clean.json

Agent

[AI Reasoning]: The error "bad escape \U at position 2" arises because the regular expression pattern includes \U, which is interpreted as a Unicode escape sequence. The problematic pattern originates from a raw string meant to match a file path, but the Python regex engine misinterprets \U. Standard string replacement, on the other hand, doesn't treat backslashes as special regex characters, so the path matching approach should work and doesn't.

Agent

Created At: 2026-10-18T19:15:29+05:30
Completed At: 2026-10-18T19:15:29+05:30
Created file file:///C:[REDACTED]/.gemini/antigravity-ide/brain/cc551c86-c432-44be-867c-b5e545626744/scratch/build_clean_session.py with requested content.
If relevant, proactively run terminal commands to execute this code for the USER. Don't ask for permission.

40 of 40 messages

My First User

Not yet :) ;)

Prize Categories

Prize Categories
I am entering the following Hacktoberfest 2026 Partner Tracks:

1. Overall 1.** Best Use of Render ** Frontend & Backend Production Hosting: Hosted entirely on Render using a FastAPI Web Service (stockmate-api) and an automated Vite React Static Site (stockmate-frontend) connected via Render custom domain routing.
2. ** Best Use of Gemma ** Core Intelligence: Google Gemma is the core intelligence engine across the system — utilizing Gemma 2:2b for local offline inference and Gemma 4:26b / Gemma 2:9b for cloud-hosted natural language parsing and grounded business Q&A.
3. ** Best Use of MongoDB Atlas ** Resilient Data Layer: Powered by MongoDB Atlas for atomic stock transactions, real-time inventory quantity adjustments ($inc), and immutable AI privacy audit logging.
4. ** Best Use of a .tech Domain ** Custom Production Domain: Deployed live on a claimed Hacktoberfest .tech domain: Application: https://stockmate.chiranjib.tech Backend API: https://api.stockmate.chiranjib.tech
5. Best Use of GitHub: Automated CI/CD with GitHub Actions (.github/workflows/ci.yml) to validate backend endpoints, run syntax checks, and test the production Vite build on every push.