Message Rewrite

v2 does not have a separate “message rewrite” table. Message-oriented rewriting is expressed through the provider rule-set system:

• system_text inserts text into the provider-native system location.
• transform replaces text patterns over the serialized body or matched paths.
• rewrite edits explicit JSON paths when you know the provider-native shape.

These rules run after protocol transform. That matters: a request from an OpenAI client routed to Claude is first transformed to Claude Messages, then message rules see the Claude body shape.

system_text

Use system_text to prepend or append server-managed instructions:

{
  "text": "Follow the internal safety policy for this workspace.",
  "position": "prepend"
}

Supported positions are prepend and append. The runtime maps the insertion to the selected content-generation kind:

Target kind	Native location
claude_messages	system string or system[] text block.
open_ai_chat_completions	A messages[] item with role: "system".
open_ai_responses	instructions.
gemini_generate_content	systemInstruction.parts[].

This is one of the few current rule kinds that knows protocol semantics. The v2 design preference is to move this kind of provider-specific path choice into frontend/config presets once the generic transform engine exists.

transform

Use transform for regex replacement when the exact structural path is not the right model:

{
  "phase": "request",
  "locate": { "match": "\\bAcme internal\\b" },
  "actions": [{ "op": "replace_text", "with": "the workspace" }]
}

The replacement runs over the serialized provider-native request body. It can modify text anywhere in the body string representation. That power is useful for prompt text, but it can also affect JSON string values you did not intend to touch. Prefer word boundaries and narrow patterns.

rewrite

Use rewrite when you know the provider-native path:

{
  "path": "messages.0.content",
  "action": "set",
  "value_json": "Pinned instruction text"
}

This is exact and structural, but it is not portable across protocol kinds. A Claude system path, an OpenAI Chat system message, an OpenAI Responses instructions string, and a Gemini systemInstruction object are different structures.

Scope Rules by Operation

Message rewrite rules should usually filter on content-generation operations:

["generate_content", "stream_generate_content"]

Avoid provider-family filters as the organizing concept. v2 classifies behavior by Operation and OperationGroup; the provider or protocol kind is a wire shape used inside that operation.

Caching Interaction

Claude prompt cache keys depend on exact prefix content. If a message rewrite changes text before a cache_control breakpoint, it can turn every request into a cache miss. Put stable cache breakpoints after rewritten content, or keep rewrite rules outside the cached prefix.