SoloBooks AI
A local-first open-source AI financial assistant for freelancers and small business owners. It uses local large models to process financial documents, and data does not leave the user's device.
Timeline (last 90 days)
SoloBooks AI is a local-first open-source finance assistant in which all financial documents never leave the user's own hard drive.
SoloBooks AI's frontend uses Next.js 16 App Router, React 19, and TypeScript strict mode.
SoloBooks AI uses SQLite as its data layer, managed via @libsql/client, achieving an embedded database with zero native builds across platforms.
SoloBooks AI stores all monetary amounts as integer minor units (e.g., $49.99 stored as 4999 cents) to eliminate floating-point precision drift.
Receipts uploaded to SoloBooks AI undergo a 15MB size limit, a MIME type whitelist, and path-traversal stripping validation before being stored under the ./data/documents/ directory with random UUID names, with the database recording only the relative path.
SoloBooks AI uses Tesseract.js v7 to OCR scanned documents and photos, and uses PDF-Parse to directly extract the native text stream of digital PDFs.
SoloBooks AI performs AI extraction through the local Ollama API (127.0.0.1:11434), having validated two models: llama3:latest (8.03B parameters) and qwen2.5:3b (3.09B parameters).
SoloBooks AI's AI extraction results are validated through a strict Zod schema (ExtractedLlmSchema), with missing fields kept as null rather than letting the model fabricate them.
SoloBooks AI adopts a human-in-the-loop design: AI extraction results are treated as draft proposals, and the user must review, edit, and confirm each item before it is written to SQLite.
When answering questions such as 'How much did I spend on software this month?', SoloBooks AI's built-in Q&A assistant first runs a deterministic SQL query, then presents verified transaction cards together with a narrative summary.
5 more timeline events
All Facts (15)
SoloBooks AI is a local-first open-source finance assistant in which all financial documents never leave the user's own hard drive.
50%UnverifiedSoloBooks AI's frontend uses Next.js 16 App Router, React 19, and TypeScript strict mode.
50%UnverifiedSoloBooks AI uses SQLite as its data layer, managed via @libsql/client, achieving an embedded database with zero native builds across platforms.
50%UnverifiedSoloBooks AI stores all monetary amounts as integer minor units (e.g., $49.99 stored as 4999 cents) to eliminate floating-point precision drift.
50%UnverifiedReceipts uploaded to SoloBooks AI undergo a 15MB size limit, a MIME type whitelist, and path-traversal stripping validation before being stored under the ./data/documents/ directory with random UUID names, with the database recording only the relative path.
50%UnverifiedSoloBooks AI uses Tesseract.js v7 to OCR scanned documents and photos, and uses PDF-Parse to directly extract the native text stream of digital PDFs.
50%UnverifiedSoloBooks AI performs AI extraction through the local Ollama API (127.0.0.1:11434), having validated two models: llama3:latest (8.03B parameters) and qwen2.5:3b (3.09B parameters).
50%UnverifiedSoloBooks AI's AI extraction results are validated through a strict Zod schema (ExtractedLlmSchema), with missing fields kept as null rather than letting the model fabricate them.
50%UnverifiedSoloBooks AI adopts a human-in-the-loop design: AI extraction results are treated as draft proposals, and the user must review, edit, and confirm each item before it is written to SQLite.
50%UnverifiedWhen answering questions such as 'How much did I spend on software this month?', SoloBooks AI's built-in Q&A assistant first runs a deterministic SQL query, then presents verified transaction cards together with a narrative summary.
50%UnverifiedSoloBooks AI's test suite runs on Vitest, with all 17 tests passing, covering minor-unit arithmetic, document security, schema validation, duplicate detection, invoice status transitions, and real Ollama connectivity.
50%UnverifiedSoloBooks AI's local model weights take up 1.9GB (3B) to 4.7GB (8B) of disk space, with the 3B model requiring at least 8GB of memory and the 8B model requiring 16GB of memory or dedicated VRAM.
50%UnverifiedOn pure CPU hardware, SoloBooks AI's OCR plus inference may take 5 to 15 seconds per document.
50%UnverifiedSoloBooks AI's current open-source core includes unlimited local OCR and PDF extraction, Ollama integration and schema validation, expense and invoice tracking, a grounded AI assistant, and CSV/JSON export.
50%UnverifiedSoloBooks AI's planned Pro tier targets a price of ₹499 per month (about $6), including intelligent mapping of bank statement CSVs, batch queue processing of up to 50 documents, password-encrypted local backup snapshots, and single-binary desktop packaging based on Tauri.
50%