vikingowl
d37cc2dad3
Extends the Content discriminated union with a fifth variant for inline image payloads. Image carries the raw bytes (captured at user-input time so the message snapshot is self-contained and survives source-file deletion), the IANA media type for the provider's image part, and the original path for logging. HasImages() lets providers decide whether to fall back to a text-only representation; providers that don't know about ContentImage will simply skip those blocks via TextContent().
102 lines
2.1 KiB
Go
102 lines
2.1 KiB
Go
package message
|
|
|
|
import "strings"
|
|
|
|
// Role identifies the sender of a message.
|
|
type Role string
|
|
|
|
const (
|
|
RoleUser Role = "user"
|
|
RoleAssistant Role = "assistant"
|
|
RoleSystem Role = "system"
|
|
)
|
|
|
|
// Message represents a single turn in the conversation.
|
|
type Message struct {
|
|
Role Role `json:"role"`
|
|
Content []Content `json:"content"`
|
|
}
|
|
|
|
func NewUserText(text string) Message {
|
|
return Message{
|
|
Role: RoleUser,
|
|
Content: []Content{NewTextContent(text)},
|
|
}
|
|
}
|
|
|
|
func NewAssistantText(text string) Message {
|
|
return Message{
|
|
Role: RoleAssistant,
|
|
Content: []Content{NewTextContent(text)},
|
|
}
|
|
}
|
|
|
|
func NewAssistantContent(blocks ...Content) Message {
|
|
return Message{
|
|
Role: RoleAssistant,
|
|
Content: blocks,
|
|
}
|
|
}
|
|
|
|
func NewSystemText(text string) Message {
|
|
return Message{
|
|
Role: RoleSystem,
|
|
Content: []Content{NewTextContent(text)},
|
|
}
|
|
}
|
|
|
|
func NewToolResults(results ...ToolResult) Message {
|
|
content := make([]Content, len(results))
|
|
for i, r := range results {
|
|
content[i] = NewToolResultContent(r)
|
|
}
|
|
return Message{
|
|
Role: RoleUser,
|
|
Content: content,
|
|
}
|
|
}
|
|
|
|
// HasToolCalls returns true if any content block is a tool call.
|
|
func (m Message) HasToolCalls() bool {
|
|
for _, c := range m.Content {
|
|
if c.Type == ContentToolCall {
|
|
return true
|
|
}
|
|
}
|
|
return false
|
|
}
|
|
|
|
// ToolCalls extracts all tool call blocks.
|
|
func (m Message) ToolCalls() []ToolCall {
|
|
var calls []ToolCall
|
|
for _, c := range m.Content {
|
|
if c.Type == ContentToolCall && c.ToolCall != nil {
|
|
calls = append(calls, *c.ToolCall)
|
|
}
|
|
}
|
|
return calls
|
|
}
|
|
|
|
// TextContent concatenates all text blocks.
|
|
func (m Message) TextContent() string {
|
|
var b strings.Builder
|
|
for _, c := range m.Content {
|
|
if c.Type == ContentText {
|
|
b.WriteString(c.Text)
|
|
}
|
|
}
|
|
return b.String()
|
|
}
|
|
|
|
// HasImages reports whether any content block in the message is an inline
|
|
// image. Providers that don't support vision can use this to decide whether
|
|
// to fall back to a text-only representation.
|
|
func (m Message) HasImages() bool {
|
|
for _, c := range m.Content {
|
|
if c.Type == ContentImage {
|
|
return true
|
|
}
|
|
}
|
|
return false
|
|
}
|