Intelligent Flask Applications

A curated collection of production-grade Flask architectures, AI agent workflows, real-time computer vision hubs, and asynchronous worker topologies.

CogniFlask AI Assistant
Production Deployed
ai-apps

CogniFlask AI Assistant

Real-time Multi-modal LLM Assistant with SSE Streaming & RAG

An enterprise-ready AI orchestration platform built on Flask 3.1, LangChain, and ChromaDB. Streams token-by-token LLM completions via Server-Sent Events (SSE), supports dynamic document ingestion, and implements semantic caching with Redis for 60% faster repeated queries.

Architectural Highlights
  • Server-Sent Events (SSE) token streaming without WebSocket overhead
  • Hybrid lexical + vector similarity search with reranking
  • Role-based access control (RBAC) and rate-limiting middleware
  • Distributed session state managed via Redis Sentinel
42ms TTFT
latency
450 req/sec
throughput
98.4% Retrieval
accuracy
1.4k
stars
Flask 3.1 LangChain ChromaDB SSE Streaming Redis Python 3.12
PromptCraft Telemetry & Evaluation
Active Service
ai-apps

PromptCraft Telemetry & Evaluation

LLM Prompt Versioning, Cost Telemetry & Automated Guardrails

Developer platform for testing, evaluating, and deploying robust LLM prompts across OpenAI, Anthropic, and local Ollama instances. Features token cost calculation, latency regression tracking, and automated toxicity filters.

Architectural Highlights
  • A/B prompt experimentation engine with semantic clustering
  • Deterministic golden dataset regression suites
  • Strict JSON schema enforcement with Pydantic v2 validation
  • Fine-grained API key usage quota limits and billing metrics
38% Token Cost
cost_saved
100 test runs in 4s
eval_speed
18+ LLM Backends
models
Flask SQLAlchemy PostgreSQL Pydantic Ollama Chart.js
DocuMind Enterprise RAG
Enterprise RAG
ai-apps

DocuMind Enterprise RAG

Production-grade Document Intelligence & Citation Engine

Specialized Flask web service parsing complex multi-page PDFs, schematics, and financial tables. Extracts structural tables, runs hybrid BM25 + dense embedding indexing, and provides verified source-highlighted answers.

Architectural Highlights
  • Pixel-accurate document bounding-box citation visualizer
  • Recursive chunking with context-aware semantic boundaries
  • Multi-tenant vector namespace isolation
  • Exportable audit reports with full grounding telemetry
250 pages / min
doc_speed
99.7%
citation_acc
PDF, DOCX, CSV, XLSX
supported_types
Flask LlamaIndex Qdrant Unstructured Tailwind CSS