NodeExacuteApi

Block execution engine for NodeNestor/AiDesigner

Built September 2023 -- February 2024, 51 commits.

What it does

REST API that executes node-based visual programs. Loads program graphs from a database, resolves block connections, executes blocks in dependency order, and returns results. This is the backend engine that powers AiDesigner.

Block types

• LLMs -- Llama 2 (7B, 13B, 70B), Zephyr 7B, Mixtral 8x7B, CodeLlama 34B
• Image AI -- Stable Diffusion XL, DALL-E 3 XL, image classification, segmentation, captioning, object detection/cropping, OCR
• Audio AI -- Whisper transcription, text-to-speech, music generation, speech enhancement
• Summarization -- BART, Flan-T5, factual consistency checking
• OpenAI -- GPT-3.5/GPT-4 integration
• Math -- arithmetic, rounding, trigonometry, random, min/max, absolute value
• String -- concatenation, split, replace, substring, case conversion, regex, length, contains, trim
• Boolean -- AND, OR, NOT, XOR
• Comparison -- equal, not equal, greater/less than, between, null check
• List -- add, remove, get, set, count, sort, filter, contains, reverse, flatten, zip, distinct
• Image manipulation -- resize, crop, rotate, flip, overlay, brightness, contrast, grayscale, blur, color filter
• Sound -- volume, trim, concatenate, reverse, speed, fade, mix
• Control flow -- if/else, switch, for loop, while loop, index loop

Token billing

Each AI model call deducts tokens from the user's account based on model-specific pricing multipliers.

Tech stack

• C# / ASP.NET Core
• SQL Server with Dapper
• HuggingFace Inference API
• OpenAI API
• TOTP-based test authentication

Status

Archived -- no longer maintained.

License

MIT