Downloads · 30 days
0
OktoSeek/oktoscript
oktoscript is a machine learning model from OktoSeek. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
<p align="center" <img src="./assets/oktologo.png" alt="OktoScript Banner" width="50%" / </p <p align="center" <img src="./assets/oktologo2.png" alt="OktoScript Banner" width="50%" / </p
Downloads · 30 days
0
Access
Public
Updated Nov 27, 2025
Repo size
6.6 MB
Likes
1
Public
Click a slice to open those files.
.png6.6 MB · 97%
From the Hugging Face model README
New to OktoScript? Get started in 5 minutes:
docs/GETTING_STARTED.mdexamples/basic.oktokto validate examples/basic.oktokto train examples/basic.okt📚 Full documentation: docs/grammar.md
🔍 Validation rules: VALIDATION_RULES.md
OktoScript is a decision-driven language created by OktoSeek AI to design, train, evaluate, control and govern AI models end-to-end.
It goes far beyond a simple training script. OktoScript introduces native intelligence, autonomous decision-making and behavioral control into the AI development lifecycle.
It allows you to define:
All using clear, readable and structured commands, built specifically for AI engineering.
OktoScript is the official language of the OktoSeek ecosystem and is used by:
Traditional AI development is reactive.
You manually monitor metrics, fix problems and restart training.
OktoScript is proactive.
It allows the model to:
In other words, OktoScript doesn't just train models — it governs intelligence.
Every OktoScript project must follow this structure:
/my-awesome-model
├── okt.yaml
├── dataset/
│ ├── train.jsonl
│ ├── val.jsonl
│ └── test.jsonl
├── scripts/
│ └── train.okt
├── runs/
│ └── my-model/
│ ├── checkpoint-100/
│ │ └── model.safetensors
│ ├── tokenizer.json
│ ├── training_logs.json
│ └── metrics.json
└── export/
├── model.gguf
├── model.onnx
└── model.okm
v1.1 Optional Folders:
/runs/
└── my-model/
├── logs/
│ └── system.json # MONITOR output (v1.1+)
└── lora/ # LoRA adapters (v1.1+)
└── adapter.safetensors
Example (v1.0 - Standard Training):
PROJECT "PizzaBot"
DESCRIPTION "AI specialized in pizza restaurant service"
ENV {
accelerator: "gpu"
min_memory: "8GB"
precision: "fp16"
backend: "oktoseek"
install_missing: true
}
DATASET {
train: "dataset/train.jsonl"
validation: "dataset/val.jsonl"
}
MODEL {
base: "oktoseek/pizza-small"
}
TRAIN {
epochs: 5
batch_size: 32
device: "auto"
}
EXPORT {
format: ["gguf", "onnx", "okm"]
path: "export/"
}
Example (v1.1 - LoRA Fine-tuning with Dataset Mixing):
# okto_version: "1.1"
PROJECT "PizzaBot"
DESCRIPTION "AI specialized in pizza restaurant service"
ENV {
accelerator: "gpu"
min_memory: "8GB"
precision: "fp16"
backend: "oktoseek"
install_missing: true
}
DATASET {
mix_datasets: [
{ path: "dataset/base.jsonl", weight: 70 },
{ path: "dataset/extra.jsonl", weight: 30 }
]
dataset_percent: 80
sampling: "weighted"
}
MODEL {
base: "oktoseek/pizza-small"
}
FT_LORA {
base_model: "oktoseek/pizza-small"
lora_rank: 8
lora_alpha: 32
epochs: 3
batch_size: 16
learning_rate: 0.00003
device: "auto"
}
MONITOR {
level: "full"
log_metrics: ["loss", "accuracy"]
log_system: ["gpu_memory_used", "cpu_usage"]
refresh_interval: 2s
dashboard: true
}
EXPORT {
format: ["okm", "onnx"]
path: "export/"
}
📘 Full grammar specification available in /docs/grammar.md
OktoScript v1.2 adds powerful new features while maintaining 100% backward compatibility with v1.0 and v1.1:
mode and prompt_style for better controldetect_using and additional prevention typeshost, protocol, and format optionsOktoScript v1.1 adds powerful new features while maintaining 100% backward compatibility with v1.0:
FT_LORA blockMONITOR block📚 More examples and use cases: See /examples/ for complete examples including:
Basic Examples:
basic.okt - Minimal examplechatbot.okt - Conversational AIcomputer_vision.okt - Image classificationrecommender.okt - Recommendation systemsAdvanced Examples:
finetuning-llm.okt - Fine-tuning LLM with checkpoints and hooksvision-pipeline.okt - Complete vision pipeline with augmentationqa-embeddings.okt - QA system with embeddingsv1.1 Examples:
lora-finetuning.okt - LoRA fine-tuning with dataset mixingdataset-mixing.okt - Training with multiple weighted datasetsComplete Projects:
pizzabot/ - Complete project example with full structureinput_field and output_field for any column names{"input":"What flavors do you have?","output":"We offer Margherita, Pepperoni and Four Cheese."}
{"input":"Do you deliver?","output":"Yes, delivery is available in your region."}
OktoScript now supports custom field names in datasets, allowing you to work with any column names:
DATASET {
train: "dataset/train.jsonl"
input_field: "question" # Custom input column name
output_field: "answer" # Custom output column name
}
If not specified, OktoEngine automatically detects input/output or input/target fields.
📖 Learn more about custom fields →
METRICS {
custom "toxicity_score"
custom "context_alignment"
}
The OktoEngine CLI is minimal by design. All intelligence lives in the .okt file. The terminal is just the execution port.
Open OktoScript files in the web editor:
# Open editor with a specific file
okto web --file scripts/train.okt
# Open empty editor
okto web
The okto web command opens the OktoScript Web Editor in your browser. When you provide a file path, it automatically loads the file content for editing. The editor features:
Perfect for quick edits, syntax testing, and experimenting with OktoScript configurations!
Initialize a project:
okto init
Validate syntax:
okto validate script.okt
Train a model:
okto train script.okt
Evaluate a model:
okto eval script.okt
Export model:
okto export script.okt
Convert model formats:
okto convert --input <model_path> --from <format> --to <format> --output <output_path>
Supported formats:
| From / To | Usage |
|---|---|
pt, bin | PyTorch |
onnx | Web / Interoperability |
tflite | Mobile (Android / iOS) |
gguf | Local LLMs (llama.cpp) |
okm | Okto Model Format |
safetensors | Safe and fast |
Convert examples:
# PyTorch → GGUF (local inference)
okto convert --input model.pt --from pt --to gguf --output model.gguf
# PyTorch → TFLite (mobile)
okto convert --input model.pt --from pt --to tflite --output model.tflite
# PyTorch → ONNX (web)
okto convert --input model.pt --from pt --to onnx --output model.onnx
List resources:
okto list projects
okto list models
okto list datasets
okto list exports
System diagnostics:
okto doctor
# Shows: GPU, CUDA, RAM, Drivers, Disks, Recommendations
Direct inference (single input/output):
okto infer --model <model_path> --text "<input>"
Example:
okto infer --model models/pizzabot.okm --text "Good evening, I want a pizza"
Automatically respects:
BEHAVIOR blockGUARD blockINFERENCE blockCONTROL block (if defined)Interactive chat mode:
okto chat --model <model_path>
Opens an interactive loop:
🟢 Okto Chat started (type 'exit' to quit)
You: hi
Bot: Hello! How can I help you?
You: what flavors do you have?
Bot: We have...
You: exit
🔴 Session ended
This command:
prompt_style from BEHAVIORBEHAVIOR settingsGUARD rulesCompare two models:
okto compare <model1> <model2>
Example:
okto compare models/pizza_v1.okm models/pizza_v2.okm
Expected output:
Latency: V2 - 23% faster
Accuracy: V1 - 4% better
Loss: V2 - lower
Recommendation: V2
Perfect for A/B testing.
View historical logs:
okto logs <model_or_run_id>
Example:
okto logs pizzabot_v1
Shows:
Auto-tune training:
okto tune script.okt
Uses the CONTROL block to auto-adjust training based on metrics. Can:
This is unique in the market.
Exit interactive mode:
okto exit
Used to exit chat, interactive mode, or session context.
okto upgrade # Update OktoEngine
okto about # Show about information
okto --version # Show version
# Validate and train
okto validate examples/basic.okt
okto train examples/chatbot.okt
# Evaluate and export
okto eval examples/recommender.okt
okto export examples/computer_vision.okt
# Inference
okto infer --model models/bot.okm --text "Hello"
okto chat --model models/bot.okm
Each run generates logs at:
runs/my-model/training_logs.json
runs/my-model/metrics.json
| Format | Purpose | Compatibility |
|---|---|---|
.onnx | Universal inference, production-ready | All platforms |
.gguf | Local inference, Ollama, Llama.cpp | Local deployment |
.safetensors | HuggingFace, research, training | Standard ML tools |
.tflite | Mobile deployment | Android, iOS (future) |
| Format | Purpose | Benefits |
|---|---|---|
.okm | OktoModel - Optimized for OktoSeek SDK | Flutter plugins, mobile apps, exclusive tools |
.okx | OktoBundle - Mobile + Edge package | iOS, Android, Edge AI deployment |
💡 Note:
.okmand.okxformats are optional and optimized for the OktoSeek ecosystem. They provide better integration with OktoSeek Flutter SDK, mobile apps, and exclusive tools. You can always export to standard formats (ONNX, GGUF, SafeTensors) for universal compatibility.
Why use OktoModel (.okm)?
See /examples/ for examples using different export formats.
Official OktoScript extension for Visual Studio Code is now available!
.okt files using OktoEngine directly from VS CodeMODEL { })From VS Code Marketplace:
Ctrl+Shift+X (or Cmd+Shift+X on Mac) to open ExtensionsOr use command line:
code --install-extension OktoSeekAI.oktoscript
Direct Link: Install OktoScript Extension
.okt file and enjoy beautiful syntax highlightingMODEL, TRAIN) and see contextual suggestionsTab to insert complete templatesCtrl+Shift+P → "OktoScript: Validate current file" (requires OktoEngine)Ctrl+Shift+P → "OktoScript: Open in Web Editor" (requires OktoEngine)💡 Tip: The VS Code extension works seamlessly with the 🌐 OktoScript Web Editor. Both provide context-aware autocomplete, real-time syntax validation, and full integration with OktoEngine via the
okto webcommand!
Complete documentation for OktoScript:
finetuning-llm.okt - Fine-tuning with checkpointsvision-pipeline.okt - Production vision systemsqa-embeddings.okt - Semantic search and retrievallora-finetuning.okt - LoRA fine-tuning (v1.1)dataset-mixing.okt - Dataset mixing (v1.1)Have questions about OktoScript? Check out our comprehensive FAQ covering common questions from beginners to advanced users:
Common Questions:
The FAQ covers technical details, design decisions, use cases, and best practices for using OktoScript effectively.
"Knowledge must be shared between people so that we can create solutions we could never imagine."
— OktoSeek AI
OktoScript is built on the principle that AI development should be:
The language evolves to support increasingly sophisticated AI behaviors while maintaining its core simplicity.
OktoScript is developed and maintained by OktoSeek AI.
OktoScript is available for personal and commercial use at no cost.
However, OktoScript is a proprietary language owned by OktoSeek AI and may not be modified or used to create derivative languages, tools or interpreters.
See OKTOSCRIPT_LICENSE.md for complete license terms.
Contributions are welcome! We welcome bug reports, feature suggestions, documentation improvements, and example contributions. Please see CONTRIBUTING.md for guidelines.
Note: OktoScript is a proprietary language owned by OktoSeek AI. While we welcome contributions, you may not create derivative languages, tools, or interpreters based on OktoScript.
If you have any questions, please raise an issue or contact us at service@oktoseek.com.