Documentation
¶
Overview ¶
Package markdowntransform holds the display-only Markdown transformer contract and its chain. It is a dependency-free leaf, so both the extension API (which re-exports these types) and the tui message components that take transformers share one definition.
Ports packages/coding-agent/src/modes/interactive/components/markdown-transform.ts
Index ¶
- func ApplyMarkdownTransformers(markdown string, ctx MarkdownTransformContext, ...) string
- func CreateMarkdownTransform(messageType MarkdownMessageType, isStreaming bool, ...) func(markdown string, availableWidth int) string
- type MarkdownMessageType
- type MarkdownTransformContext
- type MarkdownTransformer
- type RetryRequester
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func ApplyMarkdownTransformers ¶
func ApplyMarkdownTransformers(markdown string, ctx MarkdownTransformContext, transformers []MarkdownTransformer) string
ApplyMarkdownTransformers runs transformers in order on markdown. A transformer that panics keeps the current Markdown, and a done ctx.Context stops the chain with the Markdown transformed so far.
func CreateMarkdownTransform ¶
func CreateMarkdownTransform(messageType MarkdownMessageType, isStreaming bool, transformers []MarkdownTransformer, messageID, timestamp string) func(markdown string, availableWidth int) string
CreateMarkdownTransform returns the transform a Markdown component applies at its available width: each transformer rewrites the Markdown in turn, and one that panics is skipped with the current Markdown kept. It is Pi's createMarkdownTransform(messageType, isStreaming, transformers, messageId, timestamp): messageID and timestamp are the persisted entry's, or empty for a live message.
Types ¶
type MarkdownMessageType ¶
type MarkdownMessageType string
MarkdownMessageType identifies the transcript message being transformed for display. Transformers never change model context or persisted message data.
const ( MarkdownMessageUser MarkdownMessageType = "user" MarkdownMessageAssistant MarkdownMessageType = "assistant" // MarkdownMessageAssistantThinking marks the reasoning/thinking trace of an assistant turn. Mirrors upstream // MarkdownTransformContext.messageType "assistant-thinking"; the built-in Mermaid transformer skips it. MarkdownMessageAssistantThinking MarkdownMessageType = "assistant-thinking" )
type MarkdownTransformContext ¶
type MarkdownTransformContext struct {
// Context owns the off-loop host generation; it is not part of Pi's wire context.
Context context.Context `json:"-"`
// Retry is nil where nothing re-renders a message; otherwise a host proxy calls its RequestRetry, off the UI loop, once an abandoned transform can be asked again (D56). It is not part of Pi's wire context.
Retry RetryRequester `json:"-"`
MessageType MarkdownMessageType `json:"messageType"`
IsStreaming bool `json:"isStreaming"`
AvailableWidth int `json:"availableWidth"`
// MessageID is the persisted session-entry id. It is empty (Pi's undefined) for a live message or live rendering, and set when rendering from persisted session entries.
MessageID string `json:"messageId,omitempty"`
// Timestamp is the persisted session-entry ISO timestamp, with the same availability as MessageID.
Timestamp string `json:"timestamp,omitempty"`
}
MarkdownTransformContext mirrors upstream MarkdownTransformContext.
type MarkdownTransformer ¶
type MarkdownTransformer func(markdown string, context MarkdownTransformContext) string
MarkdownTransformer performs a synchronous display-only Markdown rewrite.
type RetryRequester ¶
type RetryRequester interface{ RequestRetry() }
RetryRequester marks the message being transformed as changed, so its next frame transforms it again.