solvedesk-ai 0.2.8

Local RAG framework for building intelligent knowledge bases

SolveDesk AI

Intelligent knowledge base powered by embeddings, vector search and Retrieval-Augmented Generation (RAG).

Overview

SolveDesk AI is a lightweight open-source framework for building local intelligent knowledge bases. The project provides a command-line interface for creating vector databases, importing documents, generating embeddings, performing semantic search, and integrating with local Large Language Models (LLMs).

Inspired by frameworks such as Laravel and Django, SolveDesk AI simplifies the process of building Retrieval-Augmented Generation (RAG) systems by providing ready-to-use commands and a modular architecture.

The framework can be used both as a production-ready knowledge base solution and as an educational platform for learning modern AI technologies, vector databases, embeddings, and semantic retrieval.

---

Features

• Local knowledge base creation
• Semantic document search
• Retrieval-Augmented Generation (RAG)
• Vector database management
• Embedding generation
• Local LLM integration through Ollama
• Data synchronization from APIs
• CSV, JSON and XLSX import support
• Embedding quality analysis
• Document chunking
• FastAPI integration
• Command-line interface

---

Architecture

Documents / API
       │
       ▼
Embedding Model
       │
       ▼
    ChromaDB
       │
       ▼
Semantic Search
       │
       ▼
      LLM
    Ollama
       │
       ▼
 Generated Response

---

Technologies

Technology	Purpose
Python 3.11	Application runtime
FastAPI	REST API
ChromaDB	Vector database
silver-retriever-base-v1	Embedding model
Sentence Transformers	Embedding generation
Ollama	Local LLM integration
Matplotlib	Data visualization
Typer	Command-line interface

---

Installation

Install solvedesk:

venv\Scripts\activate 
(venv) pip install solvedesk-ai

Initialize project:

(venv) C:\path\to\project> solvedesk conf init

[INFO] SolveDesk AI - Project Generator

[INPUT] Project name: Test123
[INPUT] Project description [Local RAG knowledge base]:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[INFO] Configuration
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[DETAILS] Name        : Test123
[DETAILS] Description : Local RAG knowledge base
[DETAILS] Template    : https://github.com/studiocyfrowe/solvedesk-ai

[CONFIRM] Continue project creation? [y/N]: y

[STATUS] Downloading template...

[STATUS] Project created successfully

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[INFO] Project information
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[DETAILS] Location : C:\path\to\project\Test123
[DETAILS] Name     : Test123
[DETAILS] Description : Local RAG knowledge base

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[INFO] Next steps:

cd Test123
solvedesk db init
solvedesk llm init
solvedesk run:app

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[STATUS] Happy coding!
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Initialize vector database:

(venv) C:\path\to\project> solvedesk db init

Downloading embedding model...
Model downloaded: infrastructure\models\silver-retriever-base-v1
Plik .env już istnieje — pominięto tworzenie
Created databases directory: infrastructure\databases
Created vector database: infrastructure\databases\default-db
Created/downloaded collection: random-text
SolveDesk vector DB initialized


(venv) C:\path\to\project> solvedesk db new
[STATUS] Created new collection: sd-collection-8780

(venv) C:\path\to\project> solvedesk db new --collection-name test_col
[STATUS] Created new collection: test_col

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[NEXT STEP] Show details: solvedesk db details <collection-name>
[NEXT STEP] Show list: solvedesk db list


(venv) C:\path\to\project> solvedesk db list

test_col | id=235c239e-421b-4b09-95d2-8a81bbafffd3 | documents=0 | metadata={'hnsw:space': 'cosine'}
sd-collection-8780 | id=d857b0a3-27cc-4a67-8463-4d4d075b00dd | documents=0 | metadata={'hnsw:space': 'cosine'}


(venv) C:\path\to\project> solvedesk db init --chroma-dir test12345

[CONFIRM] Download embedding model (ipipan/silver-retriever-v1)? [y/N]: n
Plik .env już istnieje — pominięto tworzenie

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[STATUS] Created databases directory: utils\databases
[STATUS] Created vector database: utils\databases\test12345
[STATUS] Vector Database is ready!

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[NEXT STEP] Create new collection: solvedesk db new <collection-name>

Configure local LLM:

solvedesk llm init

---

CLI Commands

Project Configuration

solvedesk conf init

Initialize project environment.

Database

solvedesk db init

Create vector database and download embedding model.

solvedesk db list

Display available collections.

solvedesk db details COLLECTION_NAME

Display collection details.

Data Synchronization

solvedesk sync api

Import documents from external API.

solvedesk sync file

Import documents from CSV, JSON or XLSX files.

Data Analysis

solvedesk data revision

Generate reports containing:

• cosine similarity statistics
• cluster distribution
• token statistics
• PCA visualization

Chunking

solvedesk data chunk

Split large documents into smaller chunks suitable for RAG systems.

LLM Configuration

solvedesk llm init

Configure Ollama host and model.

Run Application

solvedesk run:app

Start FastAPI server.

---

Supported Data Structures

FAQ

{
  "question": "How to reset password?",
  "answer": "Use reset password page."
}

Knowledge Base

{
  "name": "VPN Connection",
  "question": "Cannot connect to VPN",
  "answer": "Verify credentials and VPN client configuration."
}

---

Example Workflow

solvedesk conf init
solvedesk db init
solvedesk sync file
solvedesk data revision
solvedesk llm init
solvedesk run:app

---

Project Goals

• Build local intelligent knowledge bases
• Simplify RAG implementation
• Support AI experimentation
• Provide full control over data
• Enable local LLM deployments
• Offer educational value for learning AI technologies

---

License

MIT License