revert: "refactor: scaffolds anthropic messages structs for tracing" (#1516)

**Description**

This reverts commit ae516a1 due to a
bug found in an end user test. The fix seems straightforward, but we are
nearing the release v0.4, so reverting it is the best way to go for now

Signed-off-by: Takeshi Yoneda <[email protected]>

Author: mathetake

docs/why-not-official-sdk.md

@@ -3,325 +3,32 @@
 // The full text of the Apache license is available in the LICENSE file at
 // the root of the repo.
 
+// Package anthropic contains Anthropic API schema definitions using the official SDK types.
 package anthropic
 
-import (
-	"encoding/json"
-	"fmt"
-	"strings"
-
-	"github.com/tidwall/gjson"
-)
-
 // MessagesRequest represents a request to the Anthropic Messages API.
-// https://docs.claude.com/en/api/messages
-//
-// Note that we currently only have "passthrough-ish" translators for Anthropic,
-// so this struct only contains fields that are necessary for minimal processing
-// as well as for observability purposes on a best-effort basis.
-//
-// Notably, round trip idempotency is not guaranteed when using this struct.
-type MessagesRequest struct {
-	// Model is the model to use for the request.
-	Model string `json:"model,omitempty"`
-
-	// Messages is the list of messages in the conversation.
-	// https://docs.claude.com/en/api/messages#body-messages
-	Messages []Message `json:"messages"`
-
-	// MaxTokens is the maximum number of tokens to generate.
-	// https://docs.claude.com/en/api/messages#body-max-tokens
-	MaxTokens int `json:"max_tokens,omitempty"`
-
-	// Container identifier for reuse across requests.
-	// https://docs.claude.com/en/api/messages#body-container
-	Container *Container `json:"container,omitempty"`
-
-	// ContextManagement is the context management configuration.
-	// https://docs.claude.com/en/api/messages#body-context-management
-	ContextManagement *ContextManagement `json:"context_management,omitempty"`
-
-	// MCPServers is the list of MCP servers.
-	// https://docs.claude.com/en/api/messages#body-mcp-servers
-	MCPServers []MCPServer `json:"mcp_servers,omitempty"`
-
-	// Metadata is the metadata for the request.
-	// https://docs.claude.com/en/api/messages#body-metadata
-	Metadata *MessagesMetadata `json:"metadata,omitempty"`
-
-	// ServiceTier indicates the service tier for the request.
-	// https://docs.claude.com/en/api/messages#body-service-tier
-	ServiceTier *MessageServiceTier `json:"service_tier,omitempty"`
-
-	// StopSequences is the list of stop sequences.
-	// https://docs.claude.com/en/api/messages#body-stop-sequences
-	StopSequences []string `json:"stop_sequences,omitempty"`
-
-	// System is the system prompt to guide the model's behavior.
-	// https://docs.claude.com/en/api/messages#body-system
-	System *SystemPrompt `json:"system,omitempty"`
-
-	// Temperature controls the randomness of the output.
-	Temperature *float64 `json:"temperature,omitempty"`
-
-	// Thinking is the configuration for the model's "thinking" behavior.
-	// https://docs.claude.com/en/api/messages#body-thinking
-	Thinking *Thinking `json:"thinking,omitempty"`
-
-	// ToolChoice indicates the tool choice for the model.
-	// https://docs.claude.com/en/api/messages#body-tool-choice
-	ToolChoice *ToolChoice `json:"tool_choice,omitempty"`
-
-	// Tools is the list of tools available to the model.
-	// https://docs.claude.com/en/api/messages#body-tools
-	Tools []Tool `json:"tools,omitempty"`
+// Uses a dictionary approach to handle any JSON structure flexibly.
+type MessagesRequest map[string]any
 
-	// Stream indicates whether to stream the response.
-	Stream bool `json:"stream,omitempty"`
+// Helper methods to extract common fields from the dictionary
 
-	// TopP is the cumulative probability for nucleus sampling.
-	TopP *float64 `json:"top_p,omitempty"`
-
-	// TopK is the number of highest probability vocabulary tokens to keep for top-k-filtering.
-	TopK *int `json:"top_k,omitempty"`
-}
-
-// Message represents a single message in the Anthropic Messages API.
-// https://docs.claude.com/en/api/messages#body-messages
-type Message struct {
-	// Role is the role of the message.
-	Role MessageRole `json:"role"`
-
-	// Content is the content of the message.
-	Content MessageContent `json:"content"`
-}
-
-// MessageRole represents the role of a message in the Anthropic Messages API.
-// https://docs.claude.com/en/api/messages#body-messages-role
-type MessageRole string
-
-const (
-	MessageRoleUser      MessageRole = "user"
-	MessageRoleAssistant MessageRole = "assistant"
-)
-
-// MessageContent represents the content of a message in the Anthropic Messages API.
-// https://docs.claude.com/en/api/messages#body-messages-content
-type MessageContent struct {
-	Text  string                       // Non-empty iif this is not array content.
-	Array []MessageContentArrayElement // Non-empty iif this is array content.
-}
-
-// MessageContentArrayElement represents an element of the array content in a message.
-// https://docs.claude.com/en/api/messages#body-messages-content
-type MessageContentArrayElement struct{} // TODO when we need it for observability, etc.
-
-func (m *MessageContent) UnmarshalJSON(data []byte) error {
-	// Try to unmarshal as string first.
-	var text string
-	if err := json.Unmarshal(data, &text); err == nil {
-		m.Text = text
-		return nil
+func (m MessagesRequest) GetModel() string {
+	if model, ok := m["model"].(string); ok {
+		return model
 	}
-
-	// Try to unmarshal as array of MessageContentArrayElement.
-	var array []MessageContentArrayElement
-	if err := json.Unmarshal(data, &array); err == nil {
-		m.Array = array
-		return nil
-	}
-	return fmt.Errorf("message content must be either string or array")
-}
-
-// MessagesMetadata represents the metadata for the Anthropic Messages API request.
-// https://docs.claude.com/en/api/messages#body-metadata
-type MessagesMetadata struct {
-	// UserID is an optional user identifier for tracking purposes.
-	UserID *string `json:"user_id,omitempty"`
+	return ""
 }
 
-// MessageServiceTier represents the service tier for the Anthropic Messages API request.
-//
-// https://docs.claude.com/en/api/messages#body-service-tier
-type MessageServiceTier string
-
-const (
-	MessageServiceTierAuto         MessageServiceTier = "auto"
-	MessageServiceTierStandardOnly MessageServiceTier = "standard_only"
-)
-
-// Container represents a container identifier for reuse across requests.
-// https://docs.claude.com/en/api/messages#body-container
-type Container struct{} // TODO when we need it for observability, etc.
-
-// Tool represents a tool available to the model.
-// https://docs.claude.com/en/api/messages#body-tools
-type Tool struct{} // TODO when we need it for observability, etc.
-
-// ToolChoice represents the tool choice for the model.
-// https://docs.claude.com/en/api/messages#body-tool-choice
-type ToolChoice struct{} // TODO when we need it for observability, etc.
-
-// Thinking represents the configuration for the model's "thinking" behavior.
-// https://docs.claude.com/en/api/messages#body-thinking
-type Thinking struct{} // TODO when we need it for observability, etc.
-
-// SystemPrompt represents a system prompt to guide the model's behavior.
-// https://docs.claude.com/en/api/messages#body-system
-type SystemPrompt struct{} // TODO when we need it for observability, etc.
-
-// MCPServer represents an MCP server.
-// https://docs.claude.com/en/api/messages#body-mcp-servers
-type MCPServer struct{} // TODO when we need it for observability, etc.
-
-// ContextManagement represents the context management configuration.
-// https://docs.claude.com/en/api/messages#body-context-management
-type ContextManagement struct{} // TODO when we need it for observability, etc.
-
-// MessagesResponse represents a response from the Anthropic Messages API.
-// https://docs.claude.com/en/api/messages
-type MessagesResponse struct {
-	// ID is the unique identifier for the response.
-	// https://docs.claude.com/en/api/messages#response-id
-	ID string `json:"id"`
-	// Type is the type of the response.
-	// This is always "messages".
-	//
-	// https://docs.claude.com/en/api/messages#response-type
-	Type ConstantMessagesResponseTypeMessages `json:"type"`
-	// Role is the role of the message in the response.
-	// This is always "assistant".
-	//
-	// https://docs.claude.com/en/api/messages#response-role
-	Role ConstantMessagesResponseRoleAssistant `json:"role"`
-	// Content is the content of the message in the response.
-	// https://docs.claude.com/en/api/messages#response-content
-	Content []MessagesContentBlock `json:"content"`
-	// Model is the model used for the response.
-	// https://docs.claude.com/en/api/messages#response-model
-	Model string `json:"model"`
-	// StopReason is the reason for stopping the generation.
-	// https://docs.claude.com/en/api/messages#response-stop-reason
-	StopReason *StopReason `json:"stop_reason,omitempty"`
-	// StopSequence is the stop sequence that was encountered.
-	// https://docs.claude.com/en/api/messages#response-stop-sequence
-	StopSequence *string `json:"stop_sequence,omitempty"`
-	// Usage contains token usage information for the response.
-	// https://docs.claude.com/en/api/messages#response-usage
-	Usage *Usage `json:"usage,omitempty"`
-}
-
-// ConstantMessagesResponseTypeMessages is the constant type for MessagesResponse, which is always "messages".
-type ConstantMessagesResponseTypeMessages string
-
-// ConstantMessagesResponseRoleAssistant is the constant role for MessagesResponse, which is always "assistant".
-type ConstantMessagesResponseRoleAssistant string
-
-// MessagesContentBlock represents a block of content in the Anthropic Messages API response.
-// https://docs.claude.com/en/api/messages#response-content
-type MessagesContentBlock struct{} // TODO when we need it for observability, etc.
-
-// StopReason represents the reason for stopping the generation.
-// https://docs.claude.com/en/api/messages#response-stop-reason
-type StopReason string
-
-const (
-	StopReasonEndTurn                    StopReason = "end_turn"
-	StopReasonMaxTokens                  StopReason = "max_tokens"
-	StopReasonStopSequence               StopReason = "stop_sequence"
-	StopReasonToolUse                    StopReason = "tool_use"
-	StopReasonPauseTurn                  StopReason = "pause_turn"
-	StopReasonRefusal                    StopReason = "refusal"
-	StopReasonModelContextWindowExceeded StopReason = "model_context_window_exceeded"
-)
-
-// Usage represents token usage information for the Anthropic Messages API response.
-// https://docs.claude.com/en/api/messages#response-usage
-//
-// NOTE: all of them are float64 in the API, although they are always integers in practice.
-// However, the documentation doesn't explicitly state that they are integers in its format,
-// so we use float64 to be able to unmarshal both 1234 and 1234.0 without errors.
-type Usage struct {
-	// The number of input tokens used to create the cache entry.
-	CacheCreationInputTokens float64 `json:"cache_creation_input_tokens"`
-	// The number of input tokens read from the cache.
-	CacheReadInputTokens float64 `json:"cache_read_input_tokens"`
-	// The number of input tokens which were used.
-	InputTokens float64 `json:"input_tokens"`
-	// The number of output tokens which were used.
-	OutputTokens float64 `json:"output_tokens"`
-
-	// TODO: there are other fields that are currently not used in the project.
-}
-
-// MessagesStreamEvent represents a single event in the streaming response from the Anthropic Messages API.
-// https://docs.claude.com/en/docs/build-with-claude/streaming
-type MessagesStreamEvent struct {
-	// The type of the streaming event.
-	Type MessagesStreamEventType `json:"type"`
-	// MessageStart is present if the event type is "message_start" or "message_delta".
-	MessageStart *MessagesStreamEventMessageStart
-	// MessageDelta is present if the event type is "message_delta".
-	MessageDelta *MessagesStreamEventMessageDelta
-}
-
-// MessagesStreamEventType represents the type of a streaming event in the Anthropic Messages API.
-// https://docs.claude.com/en/docs/build-with-claude/streaming#event-types
-type MessagesStreamEventType string
-
-const (
-	MessagesStreamEventTypeMessageStart      MessagesStreamEventType = "message_start"
-	MessagesStreamEventTypeMessageDelta      MessagesStreamEventType = "message_delta"
-	MessagesStreamEventTypeMessageStop       MessagesStreamEventType = "message_stop"
-	MessagesStreamEventTypeContentBlockStart MessagesStreamEventType = "content_block_start"
-	MessagesStreamEventTypeContentBlockDelta MessagesStreamEventType = "content_block_delta"
-	MessagesStreamEventTypeContentBlockStop  MessagesStreamEventType = "content_block_stop"
-)
-
-// MessagesStreamEventMessageStart represents the message content in a "message_start".
-type MessagesStreamEventMessageStart MessagesResponse
-
-// MessagesStreamEventMessageDelta represents the message content in a "message_delta".
-//
-// Note: the definition of this event is vague in the Anthropic documentation.
-// This follows the same code from their official SDK.
-// https://github.com/anthropics/anthropic-sdk-go/blob/3a0275d6034e4eda9fbc8366d8a5d8b3a462b4cc/message.go#L2424-L2451
-type MessagesStreamEventMessageDelta struct {
-	// Delta contains the delta information for the message.
-	// This is cumulative per documentation.
-	Usage Usage                                `json:"usage"`
-	Delta MessagesStreamEventMessageDeltaDelta `json:"delta"`
-}
-
-type MessagesStreamEventMessageDeltaDelta struct {
-	StopReason   StopReason `json:"stop_reason"`
-	StopSequence *string    `json:"stop_sequence,omitempty"`
+func (m MessagesRequest) GetMaxTokens() int {
+	if maxTokens, ok := m["max_tokens"].(float64); ok {
+		return int(maxTokens)
+	}
+	return 0
 }
 
-func (m *MessagesStreamEvent) UnmarshalJSON(data []byte) error {
-	eventType := gjson.GetBytes(data, "type")
-	if !eventType.Exists() {
-		return fmt.Errorf("missing type field in stream event")
-	}
-	m.Type = MessagesStreamEventType(eventType.String())
-	switch m.Type {
-	case MessagesStreamEventTypeMessageStart:
-		messageBytes := gjson.GetBytes(data, "message")
-		r := strings.NewReader(messageBytes.Raw)
-		decoder := json.NewDecoder(r)
-		var message MessagesStreamEventMessageStart
-		if err := decoder.Decode(&message); err != nil {
-			return fmt.Errorf("failed to unmarshal message in stream event: %w", err)
-		}
-		m.MessageStart = &message
-	case MessagesStreamEventTypeMessageDelta:
-		var messageDelta MessagesStreamEventMessageDelta
-		if err := json.Unmarshal(data, &messageDelta); err != nil {
-			return fmt.Errorf("failed to unmarshal message delta in stream event: %w", err)
-		}
-		m.MessageDelta = &messageDelta
-	default:
-		// TODO: handle other event types if needed.
+func (m MessagesRequest) GetStream() bool {
+	if stream, ok := m["stream"].(bool); ok {
+		return stream
 	}
-	return nil
+	return false
 }