469 7 months ago

An (unofficial) ollama redistribution of the zeta editor model by the zed editor team

Models

View all →

Readme

Zed Zeta

This is a simple ollama distribution of the open source Zed Editor’s Zeta code completion model, which is based on Qwen2.5-Coder-7B. It is used within the Zed editor to provide full, live edit predictions and completions (not just lookahead autocomplete) given the context of recent changes made by the user of the editor. See this blog post for what this model can do.

Usage

This model is fine tuned by Zed on a very task specific dataset and as such the usual prompt you should pass to this model will look something like this:

### User Edits:

User edited "internal/parser/ast.py": ```diff @@ -102,3 +102,3 @@ class BinaryExpr: - def __init__(self, left, operator, right): + def __init__(self, left: Expression, operator: Token, right: Expression): self.left = left ```

### User Excerpt:

```internal/parser/ast.py class BinaryExpr: <|editable_region_start|> def __init__(self, left: Expression, operator: Token, right: Expression<|user_cursor_is_here|>): self.left = left self.operator = operator self.right = right class UnaryExpr: def __init__(self, operator, right): self.operator = operator <|editable_region_end|> self.right = right def accept(self, visitor): return visitor.visit_unary_expr(self) ```