Ollama Toolkit Conventionsο
This document outlines the conventions used across the Ollama Toolkit v0.1.8, following Eidosian principles of precision, clarity, and structural integrity.
Model Namesο
Recommended Modelsο
Latest recommended default models for different purposes:
General text generation:
deepseek-r1:1.5b
- Excellent balance of quality and efficiencyChat:
qwen2.5:0.5b
- Optimized for conversational flowEmbeddings:
nomic-embed-text
- Mathematical precision for vector representationsLightweight option:
tinyllama
- Minimal resource usage for constrained environments
The tag is optional and, if not provided, will default to latest
. The tag is used to identify a specific version of a model.
Model Name Formatο
name[:tag]@owner/model
Examples:
llama2
llama2:13b
llama2:70b-q4_0
stability.ai/stable-diffusion
API Versioningο
Ollama Toolkit follows semantic versioning (MAJOR.MINOR.PATCH):
Current version: 0.1.8
Version information is accessible via
ollama_toolkit.__version__
All endpoints maintain backward compatibility within the same MAJOR version
Durationsο
All durations in API responses are provided in nanoseconds for maximum precision.
Streaming Responsesο
Several endpoints support streaming responses. Control streaming behavior by setting:
{"stream": true}
or
{"stream": false}
in request JSON.
Streaming can be disabled by providing {"stream": false}
for these endpoints:
/api/generate
/api/chat
/api/pull
/api/push
Error Handlingο
All error responses follow a consistent format:
{
"error": "Descriptive error message"
}
HTTP status codes:
200: Success
400: Bad request (client error)
404: Not found
500: Server error
Python Client Parametersο
When using the Python client, parameters follow consistent naming:
Use snake_case for parameter names
Boolean flags use explicit names (e.g.,
stream=True
notstream
)Optional parameters default to Pythonβs
None
where appropriate
Documentation Styleο
All documentation follows these Eidosian principles:
Contextual Integrity: Every section serves a precise purpose
Exhaustive But Concise: Complete information presented efficiently
Flow Like a River: Smooth transitions between related concepts
Precision as Style: Clear, accurate information presented elegantly
CLI Conventionsο
Command-line interface commands use consistent patterns:
Main subcommands represent primary API functions
Options follow GNU-style with both short and long forms (e.g.,
-m/--model
)Global flags appear before subcommands, specific flags after