Ryan WardandClaude Opus 4.6
eaa2ddc75e
Major changes: - **Bedrock translator architecture**: Extract orchestration logic from `impl.rs` into a dedicated `translator.rs` module. Rename `convert_request.rs` → `request_translator.rs` and `stream.rs` → `response_translator.rs` for clarity. Remove `tool_docs.rs` (inlined). Remove `fallback_to_warp` setting and server fallback path — Bedrock is now the sole backend. - **Unknown tool handling**: The response translator now detects hallucinated/unknown tool calls from the model and synthesizes error tool_results so the conversation doesn't deadlock waiting for a result that will never come. - **Usage display overhaul**: Replace credit-based usage display with detailed token metrics showing context window %, cache hit rate (read/write/miss), and estimated cost in dollars. Add `total_input_tokens`, `total_cache_read_tokens`, `total_cache_write_tokens`, and `cache_miss_tokens` accessors to `AIConversation`. - **Predefined rules system**: Add `predefined_rules.rs` with 11 system-defined behavioral rules that are auto-seeded on first launch. Add "Add Predefined Rules" button to the Rules UI for re-adding them later. Track seeding state via `has_seeded_predefined_rules` setting. - **Session restore improvements**: Rename database file from `warp.sqlite` to `galaxy.sqlite` with automatic migration from both same-directory and state_dir legacy paths. Improve CWD persistence by falling back to `session_startup_path` for agent-mode and fresh tabs. Add extensive session-save/restore logging. - **Shell bootstrap rebrand**: Rename `WARP_INITIAL_WORKING_DIR` environment variable to `GALAXY_INITIAL_WORKING_DIR` across bash, zsh, and fish bootstrap scripts. - **Model defaults**: Change default Bedrock model from Opus 4.7 to Opus 4.6. Add `context_window_for_model()` helper with model-aware context sizes. Remove `is_bedrock_model()` (no longer needed without server fallback). - **User query persistence**: The response translator now emits a `UserQuery` proto message at stream start so the user's prompt persists across sessions for conversation titles. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2165 lines
74 KiB
Rust
2165 lines
74 KiB
Rust
use std::sync::{Arc, Mutex};
|
|
|
|
use futures::StreamExt;
|
|
use serde_json::json;
|
|
use std::path::PathBuf;
|
|
|
|
use super::client::{BedrockClient, BedrockClientConfig};
|
|
use super::convert::{ConversationMessage, MessageContent, MessageRole, ToolDefinition};
|
|
use crate::settings::ai::BedrockAuthMethod;
|
|
use warp_multi_agent_api as api;
|
|
|
|
fn make_user_query_message(id: &str, task_id: &str, query: &str) -> api::Message {
|
|
api::Message {
|
|
id: id.into(),
|
|
task_id: task_id.into(),
|
|
request_id: String::new(),
|
|
timestamp: None,
|
|
server_message_data: String::new(),
|
|
citations: vec![],
|
|
message: Some(api::message::Message::UserQuery(api::message::UserQuery {
|
|
query: query.into(),
|
|
context: None,
|
|
referenced_attachments: std::collections::HashMap::new(),
|
|
mode: None,
|
|
intended_agent: 0,
|
|
})),
|
|
}
|
|
}
|
|
|
|
fn make_tool_call_run_shell(
|
|
id: &str,
|
|
task_id: &str,
|
|
tool_call_id: &str,
|
|
command: &str,
|
|
) -> api::Message {
|
|
api::Message {
|
|
id: id.into(),
|
|
task_id: task_id.into(),
|
|
request_id: String::new(),
|
|
timestamp: None,
|
|
server_message_data: String::new(),
|
|
citations: vec![],
|
|
message: Some(api::message::Message::ToolCall(api::message::ToolCall {
|
|
tool_call_id: tool_call_id.into(),
|
|
tool: Some(api::message::tool_call::Tool::RunShellCommand(
|
|
api::message::tool_call::RunShellCommand {
|
|
command: command.into(),
|
|
is_read_only: true,
|
|
uses_pager: false,
|
|
citations: vec![],
|
|
is_risky: false,
|
|
risk_category: 0,
|
|
wait_until_complete_value: None,
|
|
},
|
|
)),
|
|
})),
|
|
}
|
|
}
|
|
|
|
fn make_tool_call_read_files(
|
|
id: &str,
|
|
task_id: &str,
|
|
tool_call_id: &str,
|
|
file_name: &str,
|
|
) -> api::Message {
|
|
api::Message {
|
|
id: id.into(),
|
|
task_id: task_id.into(),
|
|
request_id: String::new(),
|
|
timestamp: None,
|
|
server_message_data: String::new(),
|
|
citations: vec![],
|
|
message: Some(api::message::Message::ToolCall(api::message::ToolCall {
|
|
tool_call_id: tool_call_id.into(),
|
|
tool: Some(api::message::tool_call::Tool::ReadFiles(
|
|
api::message::tool_call::ReadFiles {
|
|
files: vec![api::message::tool_call::read_files::File {
|
|
name: file_name.into(),
|
|
line_ranges: vec![],
|
|
}],
|
|
},
|
|
)),
|
|
})),
|
|
}
|
|
}
|
|
|
|
#[allow(deprecated)]
|
|
fn make_tool_result_shell(
|
|
id: &str,
|
|
task_id: &str,
|
|
tool_call_id: &str,
|
|
output: &str,
|
|
exit_code: i32,
|
|
) -> api::Message {
|
|
api::Message {
|
|
id: id.into(),
|
|
task_id: task_id.into(),
|
|
request_id: String::new(),
|
|
timestamp: None,
|
|
server_message_data: String::new(),
|
|
citations: vec![],
|
|
message: Some(api::message::Message::ToolCallResult(
|
|
api::message::ToolCallResult {
|
|
tool_call_id: tool_call_id.into(),
|
|
context: None,
|
|
result: Some(api::message::tool_call_result::Result::RunShellCommand(
|
|
api::RunShellCommandResult {
|
|
command: String::new(),
|
|
output: String::new(),
|
|
exit_code: 0,
|
|
result: Some(api::run_shell_command_result::Result::CommandFinished(
|
|
api::ShellCommandFinished {
|
|
output: output.into(),
|
|
exit_code,
|
|
command_id: String::new(),
|
|
},
|
|
)),
|
|
},
|
|
)),
|
|
},
|
|
)),
|
|
}
|
|
}
|
|
|
|
fn make_tool_result_read_files(
|
|
id: &str,
|
|
task_id: &str,
|
|
tool_call_id: &str,
|
|
file_path: &str,
|
|
content: &str,
|
|
) -> api::Message {
|
|
api::Message {
|
|
id: id.into(),
|
|
task_id: task_id.into(),
|
|
request_id: String::new(),
|
|
timestamp: None,
|
|
server_message_data: String::new(),
|
|
citations: vec![],
|
|
message: Some(api::message::Message::ToolCallResult(
|
|
api::message::ToolCallResult {
|
|
tool_call_id: tool_call_id.into(),
|
|
context: None,
|
|
result: Some(api::message::tool_call_result::Result::ReadFiles(
|
|
api::ReadFilesResult {
|
|
result: Some(api::read_files_result::Result::TextFilesSuccess(
|
|
api::read_files_result::TextFilesSuccess {
|
|
files: vec![api::FileContent {
|
|
file_path: file_path.into(),
|
|
content: content.into(),
|
|
line_range: None,
|
|
}],
|
|
},
|
|
)),
|
|
},
|
|
)),
|
|
},
|
|
)),
|
|
}
|
|
}
|
|
|
|
fn make_user_inputs_input(query: &str) -> api::request::Input {
|
|
api::request::Input {
|
|
context: None,
|
|
r#type: Some(api::request::input::Type::UserInputs(
|
|
api::request::input::UserInputs {
|
|
inputs: vec![api::request::input::user_inputs::UserInput {
|
|
input: Some(
|
|
api::request::input::user_inputs::user_input::Input::UserQuery(
|
|
api::request::input::UserQuery {
|
|
query: query.into(),
|
|
referenced_attachments: std::collections::HashMap::new(),
|
|
mode: None,
|
|
intended_agent: 0,
|
|
},
|
|
),
|
|
),
|
|
}],
|
|
},
|
|
)),
|
|
}
|
|
}
|
|
|
|
#[allow(deprecated)]
|
|
fn make_settings(model: &str) -> api::request::Settings {
|
|
api::request::Settings {
|
|
model_config: Some(api::request::settings::ModelConfig {
|
|
base: model.into(),
|
|
planning: String::new(),
|
|
coding: String::new(),
|
|
cli_agent: String::new(),
|
|
computer_use_agent: String::new(),
|
|
base_model_context_window_limit: 0,
|
|
}),
|
|
rules_enabled: false,
|
|
web_context_retrieval_enabled: false,
|
|
supports_parallel_tool_calls: true,
|
|
use_anthropic_text_editor_tools: false,
|
|
planning_enabled: false,
|
|
warp_drive_context_enabled: false,
|
|
supports_create_files: true,
|
|
supported_tools: vec![],
|
|
supports_long_running_commands: false,
|
|
should_preserve_file_content_in_history: true,
|
|
supports_todos_ui: false,
|
|
supports_linked_code_blocks: false,
|
|
supports_started_child_task_message: false,
|
|
supports_suggest_prompt: false,
|
|
supports_read_image_files: false,
|
|
supports_reasoning_message: true,
|
|
api_keys: None,
|
|
autonomy_level: 0,
|
|
isolation_level: 0,
|
|
web_search_enabled: false,
|
|
supported_cli_agent_tools: vec![],
|
|
supports_v4a_file_diffs: false,
|
|
supports_summarization_via_message_replacement: false,
|
|
supports_bundled_skills: false,
|
|
supports_research_agent: false,
|
|
supports_orchestration_v2: false,
|
|
}
|
|
}
|
|
|
|
fn make_request(
|
|
task_id: &str,
|
|
messages: Vec<api::Message>,
|
|
input: api::request::Input,
|
|
model: &str,
|
|
) -> api::Request {
|
|
api::Request {
|
|
task_context: Some(api::request::TaskContext {
|
|
tasks: vec![api::Task {
|
|
id: task_id.into(),
|
|
description: String::new(),
|
|
dependencies: None,
|
|
messages,
|
|
summary: String::new(),
|
|
server_data: String::new(),
|
|
}],
|
|
}),
|
|
input: Some(input),
|
|
settings: Some(make_settings(model)),
|
|
metadata: None,
|
|
existing_suggestions: None,
|
|
mcp_context: None,
|
|
}
|
|
}
|
|
|
|
fn get_test_config() -> Option<BedrockClientConfig> {
|
|
if std::env::var("BEDROCK_INTEGRATION_TEST").is_err() {
|
|
return None;
|
|
}
|
|
|
|
let profile =
|
|
std::env::var("BEDROCK_TEST_PROFILE").unwrap_or_else(|_| "coding-assistant".into());
|
|
let region = std::env::var("BEDROCK_TEST_REGION").unwrap_or_else(|_| "us-east-1".into());
|
|
|
|
Some(BedrockClientConfig {
|
|
auth_method: BedrockAuthMethod::Profile,
|
|
profile,
|
|
region,
|
|
access_key_id: String::new(),
|
|
secret_access_key: String::new(),
|
|
cross_region_inference: false,
|
|
})
|
|
}
|
|
|
|
fn get_test_model() -> String {
|
|
std::env::var("BEDROCK_TEST_MODEL").unwrap_or_else(|_| {
|
|
"arn:aws:bedrock:us-east-1:156729649053:application-inference-profile/fma5bsw4oxhy".into()
|
|
})
|
|
}
|
|
|
|
fn sample_project_path() -> PathBuf {
|
|
PathBuf::from(env!("CARGO_MANIFEST_DIR")).join("src/ai/bedrock/test_fixtures/sample_project")
|
|
}
|
|
|
|
fn agent_tools() -> Vec<ToolDefinition> {
|
|
vec![
|
|
ToolDefinition {
|
|
name: "run_shell_command".into(),
|
|
description: "Execute a shell command and return its output.".into(),
|
|
input_schema: json!({
|
|
"type": "object",
|
|
"properties": {
|
|
"command": { "type": "string", "description": "The shell command to execute" }
|
|
},
|
|
"required": ["command"]
|
|
}),
|
|
},
|
|
ToolDefinition {
|
|
name: "read_files".into(),
|
|
description: "Read the contents of one or more files.".into(),
|
|
input_schema: json!({
|
|
"type": "object",
|
|
"properties": {
|
|
"files": { "type": "array", "items": { "type": "string" }, "description": "File paths to read" }
|
|
},
|
|
"required": ["files"]
|
|
}),
|
|
},
|
|
ToolDefinition {
|
|
name: "apply_file_diffs".into(),
|
|
description: "Apply search/replace diffs to files. Creates a file if it doesn't exist."
|
|
.into(),
|
|
input_schema: json!({
|
|
"type": "object",
|
|
"properties": {
|
|
"diffs": {
|
|
"type": "array",
|
|
"items": {
|
|
"type": "object",
|
|
"properties": {
|
|
"file_path": { "type": "string" },
|
|
"search": { "type": "string", "description": "Text to find (empty for new file)" },
|
|
"replace": { "type": "string", "description": "Replacement text" }
|
|
},
|
|
"required": ["file_path", "search", "replace"]
|
|
}
|
|
}
|
|
},
|
|
"required": ["diffs"]
|
|
}),
|
|
},
|
|
ToolDefinition {
|
|
name: "grep".into(),
|
|
description: "Search for patterns in files using grep.".into(),
|
|
input_schema: json!({
|
|
"type": "object",
|
|
"properties": {
|
|
"queries": { "type": "array", "items": { "type": "string" }, "description": "Search patterns" },
|
|
"path": { "type": "string", "description": "Directory to search in" }
|
|
},
|
|
"required": ["queries"]
|
|
}),
|
|
},
|
|
ToolDefinition {
|
|
name: "file_glob".into(),
|
|
description: "Find files matching glob patterns.".into(),
|
|
input_schema: json!({
|
|
"type": "object",
|
|
"properties": {
|
|
"patterns": { "type": "array", "items": { "type": "string" }, "description": "Glob patterns to match" }
|
|
},
|
|
"required": ["patterns"]
|
|
}),
|
|
},
|
|
]
|
|
}
|
|
|
|
#[derive(Debug)]
|
|
struct StreamEvent {
|
|
event_type: StreamEventType,
|
|
}
|
|
|
|
#[derive(Debug)]
|
|
enum StreamEventType {
|
|
Init {
|
|
request_id: String,
|
|
conversation_id: String,
|
|
},
|
|
CreateTask {
|
|
task_id: String,
|
|
},
|
|
AddMessage {
|
|
task_id: String,
|
|
message_id: String,
|
|
text: String,
|
|
},
|
|
AppendText {
|
|
task_id: String,
|
|
message_id: String,
|
|
delta: String,
|
|
},
|
|
ToolCallMessage {
|
|
task_id: String,
|
|
text: String,
|
|
},
|
|
Finished {
|
|
reason: String,
|
|
total_tokens: u32,
|
|
},
|
|
Error(String),
|
|
}
|
|
|
|
fn parse_response_event(event: &api::ResponseEvent) -> StreamEvent {
|
|
match &event.r#type {
|
|
Some(api::response_event::Type::Init(init)) => StreamEvent {
|
|
event_type: StreamEventType::Init {
|
|
request_id: init.request_id.clone(),
|
|
conversation_id: init.conversation_id.clone(),
|
|
},
|
|
},
|
|
Some(api::response_event::Type::ClientActions(actions)) => {
|
|
for action in &actions.actions {
|
|
if let Some(action_type) = &action.action {
|
|
match action_type {
|
|
api::client_action::Action::CreateTask(ct) => {
|
|
let task_id =
|
|
ct.task.as_ref().map(|t| t.id.clone()).unwrap_or_default();
|
|
return StreamEvent {
|
|
event_type: StreamEventType::CreateTask { task_id },
|
|
};
|
|
}
|
|
api::client_action::Action::AddMessagesToTask(add) => {
|
|
let task_id = add.task_id.clone();
|
|
for msg in &add.messages {
|
|
if let Some(api::message::Message::AgentOutput(output)) =
|
|
&msg.message
|
|
{
|
|
if output.text.starts_with("[Tool call:") {
|
|
return StreamEvent {
|
|
event_type: StreamEventType::ToolCallMessage {
|
|
task_id,
|
|
text: output.text.clone(),
|
|
},
|
|
};
|
|
}
|
|
return StreamEvent {
|
|
event_type: StreamEventType::AddMessage {
|
|
task_id,
|
|
message_id: msg.id.clone(),
|
|
text: output.text.clone(),
|
|
},
|
|
};
|
|
}
|
|
}
|
|
}
|
|
api::client_action::Action::AppendToMessageContent(append) => {
|
|
let task_id = append.task_id.clone();
|
|
if let Some(msg) = &append.message {
|
|
let message_id = msg.id.clone();
|
|
if let Some(api::message::Message::AgentOutput(output)) =
|
|
&msg.message
|
|
{
|
|
return StreamEvent {
|
|
event_type: StreamEventType::AppendText {
|
|
task_id,
|
|
message_id,
|
|
delta: output.text.clone(),
|
|
},
|
|
};
|
|
}
|
|
}
|
|
}
|
|
_ => {}
|
|
}
|
|
}
|
|
}
|
|
StreamEvent {
|
|
event_type: StreamEventType::Error("Unknown action".into()),
|
|
}
|
|
}
|
|
Some(api::response_event::Type::Finished(finished)) => {
|
|
let reason = format!("{:?}", finished.reason);
|
|
let total_tokens = finished
|
|
.conversation_usage_metadata
|
|
.as_ref()
|
|
.and_then(|m| m.byok_token_usage.get("bedrock"))
|
|
.map(|u| u.total_tokens)
|
|
.unwrap_or(0);
|
|
StreamEvent {
|
|
event_type: StreamEventType::Finished {
|
|
reason,
|
|
total_tokens,
|
|
},
|
|
}
|
|
}
|
|
None => StreamEvent {
|
|
event_type: StreamEventType::Error("Empty event".into()),
|
|
},
|
|
}
|
|
}
|
|
|
|
struct AgentSimulation {
|
|
client: BedrockClient,
|
|
model: String,
|
|
task_id: String,
|
|
conversation: Vec<ConversationMessage>,
|
|
tools: Vec<ToolDefinition>,
|
|
system_prompt: String,
|
|
project_path: PathBuf,
|
|
all_text_output: String,
|
|
turn_count: u32,
|
|
max_turns: u32,
|
|
}
|
|
|
|
#[derive(Debug)]
|
|
struct TurnResult {
|
|
text_output: String,
|
|
tool_calls: Vec<ToolCall>,
|
|
finished: bool,
|
|
total_tokens: u32,
|
|
events_received: u32,
|
|
had_create_task: bool,
|
|
}
|
|
|
|
#[derive(Debug, Clone)]
|
|
struct ToolCall {
|
|
name: String,
|
|
input: serde_json::Value,
|
|
raw_text: String,
|
|
}
|
|
|
|
impl AgentSimulation {
|
|
fn new(client: BedrockClient, model: String, project_path: PathBuf) -> Self {
|
|
let task_id = uuid::Uuid::new_v4().to_string();
|
|
Self {
|
|
client,
|
|
model,
|
|
task_id,
|
|
conversation: Vec::new(),
|
|
tools: agent_tools(),
|
|
system_prompt: format!(
|
|
"You are a helpful AI coding assistant. You help users with software engineering tasks. \
|
|
The user's current working directory is: {}. \
|
|
When using tools, use absolute paths based on this directory. \
|
|
When asked to produce files, use the apply_file_diffs tool to create them.",
|
|
project_path.display()
|
|
),
|
|
project_path,
|
|
all_text_output: String::new(),
|
|
turn_count: 0,
|
|
max_turns: 10,
|
|
}
|
|
}
|
|
|
|
async fn send_user_message(&mut self, message: &str) -> TurnResult {
|
|
self.conversation.push(ConversationMessage {
|
|
role: MessageRole::User,
|
|
content: MessageContent::Text(message.into()),
|
|
});
|
|
self.execute_turn().await
|
|
}
|
|
|
|
async fn execute_turn(&mut self) -> TurnResult {
|
|
self.turn_count += 1;
|
|
assert!(
|
|
self.turn_count <= self.max_turns,
|
|
"Exceeded max turns ({})",
|
|
self.max_turns
|
|
);
|
|
|
|
let needs_create_task = self.turn_count == 1;
|
|
|
|
let stream = self
|
|
.client
|
|
.converse_stream(
|
|
&self.model,
|
|
&self.task_id,
|
|
needs_create_task,
|
|
self.conversation.clone(),
|
|
Some(self.system_prompt.clone()),
|
|
self.tools.clone(),
|
|
8192,
|
|
None,
|
|
false,
|
|
None,
|
|
None,
|
|
Arc::new(Mutex::new(Vec::new())),
|
|
)
|
|
.await
|
|
.expect("converse_stream should succeed");
|
|
|
|
let mut text_output = String::new();
|
|
let mut tool_calls = Vec::new();
|
|
let mut finished = false;
|
|
let mut total_tokens = 0u32;
|
|
let mut events_received = 0u32;
|
|
let mut had_create_task = false;
|
|
let mut had_init = false;
|
|
|
|
let mut stream = stream;
|
|
while let Some(event_result) = stream.next().await {
|
|
events_received += 1;
|
|
let event = event_result.expect("stream event should be Ok");
|
|
let parsed = parse_response_event(&event);
|
|
|
|
match parsed.event_type {
|
|
StreamEventType::Init { request_id, .. } => {
|
|
had_init = true;
|
|
assert!(
|
|
!request_id.is_empty(),
|
|
"StreamInit should have non-empty request_id"
|
|
);
|
|
}
|
|
StreamEventType::CreateTask { ref task_id } => {
|
|
had_create_task = true;
|
|
assert_eq!(
|
|
task_id, &self.task_id,
|
|
"CreateTask task_id should match our task_id"
|
|
);
|
|
}
|
|
StreamEventType::AddMessage {
|
|
ref task_id,
|
|
text: ref t,
|
|
..
|
|
} => {
|
|
assert_eq!(
|
|
task_id, &self.task_id,
|
|
"AddMessage task_id should match our task_id"
|
|
);
|
|
text_output.push_str(t);
|
|
}
|
|
StreamEventType::AppendText {
|
|
ref task_id,
|
|
delta: ref d,
|
|
..
|
|
} => {
|
|
assert_eq!(
|
|
task_id, &self.task_id,
|
|
"AppendText task_id should match our task_id"
|
|
);
|
|
text_output.push_str(d);
|
|
}
|
|
StreamEventType::ToolCallMessage { text: ref t, .. } => {
|
|
if let Some(call) = parse_tool_call_text(t) {
|
|
tool_calls.push(call);
|
|
}
|
|
}
|
|
StreamEventType::Finished {
|
|
total_tokens: tokens,
|
|
..
|
|
} => {
|
|
finished = true;
|
|
total_tokens = tokens;
|
|
}
|
|
StreamEventType::Error(ref e) => {
|
|
panic!("Unexpected error event: {e}");
|
|
}
|
|
}
|
|
}
|
|
|
|
assert!(had_init, "Stream should start with Init event");
|
|
if needs_create_task {
|
|
assert!(had_create_task, "First turn should have CreateTask event");
|
|
}
|
|
assert!(finished, "Stream should end with Finished event");
|
|
assert!(
|
|
events_received >= 3,
|
|
"Should receive at least Init + one content + Finished"
|
|
);
|
|
|
|
self.all_text_output.push_str(&text_output);
|
|
|
|
if !text_output.is_empty() {
|
|
self.conversation.push(ConversationMessage {
|
|
role: MessageRole::Assistant,
|
|
content: MessageContent::Text(text_output.clone()),
|
|
});
|
|
}
|
|
|
|
TurnResult {
|
|
text_output,
|
|
tool_calls,
|
|
finished,
|
|
total_tokens,
|
|
events_received,
|
|
had_create_task,
|
|
}
|
|
}
|
|
|
|
fn provide_tool_result(
|
|
&mut self,
|
|
tool_use_id: &str,
|
|
name: &str,
|
|
input: &serde_json::Value,
|
|
result: &str,
|
|
is_error: bool,
|
|
) {
|
|
self.conversation.push(ConversationMessage {
|
|
role: MessageRole::Assistant,
|
|
content: MessageContent::ToolUse {
|
|
tool_use_id: tool_use_id.into(),
|
|
name: name.into(),
|
|
input: input.clone(),
|
|
},
|
|
});
|
|
self.conversation.push(ConversationMessage {
|
|
role: MessageRole::User,
|
|
content: MessageContent::ToolResult {
|
|
tool_use_id: tool_use_id.into(),
|
|
content: result.into(),
|
|
is_error,
|
|
},
|
|
});
|
|
}
|
|
|
|
fn execute_tool_locally(&self, name: &str, input: &serde_json::Value) -> (String, bool) {
|
|
match name {
|
|
"run_shell_command" => {
|
|
let command = input["command"].as_str().unwrap_or("echo 'no command'");
|
|
let output = std::process::Command::new("sh")
|
|
.arg("-c")
|
|
.arg(command)
|
|
.current_dir(&self.project_path)
|
|
.output();
|
|
match output {
|
|
Ok(out) => {
|
|
let stdout = String::from_utf8_lossy(&out.stdout);
|
|
let stderr = String::from_utf8_lossy(&out.stderr);
|
|
let result = if stderr.is_empty() {
|
|
stdout.to_string()
|
|
} else {
|
|
format!("{}\nstderr: {}", stdout, stderr)
|
|
};
|
|
(result, !out.status.success())
|
|
}
|
|
Err(e) => (format!("Error: {e}"), true),
|
|
}
|
|
}
|
|
"read_files" => {
|
|
let files = input["files"]
|
|
.as_array()
|
|
.map(|arr| arr.iter().filter_map(|v| v.as_str()).collect::<Vec<_>>())
|
|
.unwrap_or_default();
|
|
let mut result = String::new();
|
|
for file in files {
|
|
let path = if file.starts_with('/') {
|
|
PathBuf::from(file)
|
|
} else {
|
|
self.project_path.join(file)
|
|
};
|
|
match std::fs::read_to_string(&path) {
|
|
Ok(content) => {
|
|
result.push_str(&format!("{}:\n{}\n\n", path.display(), content));
|
|
}
|
|
Err(e) => {
|
|
result.push_str(&format!("Error reading {}: {}\n", path.display(), e));
|
|
}
|
|
}
|
|
}
|
|
(result, false)
|
|
}
|
|
"file_glob" => {
|
|
let mut files = Vec::new();
|
|
for entry in walkdir::WalkDir::new(&self.project_path)
|
|
.into_iter()
|
|
.filter_map(|e| e.ok())
|
|
.filter(|e| e.file_type().is_file())
|
|
{
|
|
files.push(entry.path().display().to_string());
|
|
}
|
|
let result = if files.is_empty() {
|
|
"No files found.".into()
|
|
} else {
|
|
files.join("\n")
|
|
};
|
|
(result, false)
|
|
}
|
|
"grep" => {
|
|
let queries = input["queries"]
|
|
.as_array()
|
|
.map(|arr| arr.iter().filter_map(|v| v.as_str()).collect::<Vec<_>>())
|
|
.unwrap_or_default();
|
|
let path = input["path"]
|
|
.as_str()
|
|
.unwrap_or(self.project_path.to_str().unwrap_or("."));
|
|
let mut result = String::new();
|
|
for query in queries {
|
|
let output = std::process::Command::new("grep")
|
|
.args(["-rn", query, path])
|
|
.output();
|
|
if let Ok(out) = output {
|
|
result.push_str(&String::from_utf8_lossy(&out.stdout));
|
|
}
|
|
}
|
|
if result.is_empty() {
|
|
result = "No matches found.".into();
|
|
}
|
|
(result, false)
|
|
}
|
|
"apply_file_diffs" => {
|
|
let diffs = input["diffs"].as_array();
|
|
match diffs {
|
|
Some(diffs) => {
|
|
let mut results = Vec::new();
|
|
for diff in diffs {
|
|
let file_path = diff["file_path"].as_str().unwrap_or("");
|
|
let search = diff["search"].as_str().unwrap_or("");
|
|
let replace = diff["replace"].as_str().unwrap_or("");
|
|
|
|
let path = if file_path.starts_with('/') {
|
|
PathBuf::from(file_path)
|
|
} else {
|
|
self.project_path.join(file_path)
|
|
};
|
|
|
|
if search.is_empty() {
|
|
if let Some(parent) = path.parent() {
|
|
let _ = std::fs::create_dir_all(parent);
|
|
}
|
|
match std::fs::write(&path, replace) {
|
|
Ok(_) => results.push(format!("Created: {}", path.display())),
|
|
Err(e) => results.push(format!(
|
|
"Error creating {}: {}",
|
|
path.display(),
|
|
e
|
|
)),
|
|
}
|
|
} else {
|
|
match std::fs::read_to_string(&path) {
|
|
Ok(content) => {
|
|
let new_content = content.replace(search, replace);
|
|
match std::fs::write(&path, &new_content) {
|
|
Ok(_) => {
|
|
results.push(format!("Updated: {}", path.display()))
|
|
}
|
|
Err(e) => results.push(format!(
|
|
"Error writing {}: {}",
|
|
path.display(),
|
|
e
|
|
)),
|
|
}
|
|
}
|
|
Err(e) => results.push(format!(
|
|
"Error reading {}: {}",
|
|
path.display(),
|
|
e
|
|
)),
|
|
}
|
|
}
|
|
}
|
|
(results.join("\n"), false)
|
|
}
|
|
None => ("No diffs provided.".into(), true),
|
|
}
|
|
}
|
|
_ => (format!("Unknown tool: {name}"), true),
|
|
}
|
|
}
|
|
}
|
|
|
|
fn parse_tool_call_text(text: &str) -> Option<ToolCall> {
|
|
if text.starts_with("[Tool call: ") && text.ends_with("]") {
|
|
let inner = &text[12..text.len() - 1];
|
|
if let Some(paren_pos) = inner.rfind(" (") {
|
|
let name = inner[..paren_pos].to_string();
|
|
return Some(ToolCall {
|
|
name,
|
|
input: json!({}),
|
|
raw_text: text.to_string(),
|
|
});
|
|
}
|
|
}
|
|
None
|
|
}
|
|
|
|
/// This test simulates the exact flow when a user types:
|
|
/// /agent Please review this code and produce a WARP.md file for it.
|
|
///
|
|
/// It verifies:
|
|
/// 1. StreamInit is emitted first
|
|
/// 2. CreateTask is emitted for new conversations (upgrades optimistic root task)
|
|
/// 3. Text output is substantial (not just "#" or empty)
|
|
/// 4. Tool calls are properly formed and can be executed
|
|
/// 5. Multi-turn conversation works (tool results fed back)
|
|
/// 6. The model eventually produces meaningful text output
|
|
/// 7. Stream finishes cleanly with token usage
|
|
#[tokio::test]
|
|
async fn test_agent_flow_review_code_produce_warp_md() {
|
|
let Some(config) = get_test_config() else {
|
|
eprintln!("Skipping: BEDROCK_INTEGRATION_TEST not set");
|
|
return;
|
|
};
|
|
|
|
let client = BedrockClient::from_config(config)
|
|
.await
|
|
.expect("client creation");
|
|
let model = get_test_model();
|
|
let project_path = sample_project_path();
|
|
|
|
println!("\n{}", "=".repeat(80));
|
|
println!("[E2E TEST] Agent flow: /agent Please review this code and produce a WARP.md file");
|
|
println!("[E2E TEST] Project: {}", project_path.display());
|
|
println!("[E2E TEST] Model: {model}");
|
|
println!("{}\n", "=".repeat(80));
|
|
|
|
let mut sim = AgentSimulation::new(client, model, project_path.clone());
|
|
|
|
// Simulate the /agent command
|
|
let turn1 = sim
|
|
.send_user_message("Please review this code and produce a WARP.md file for it.")
|
|
.await;
|
|
|
|
println!("\n--- Turn 1 ---");
|
|
println!(
|
|
"[E2E] Text ({} chars): {:?}",
|
|
turn1.text_output.len(),
|
|
&turn1.text_output[..turn1.text_output.len().min(500)]
|
|
);
|
|
println!("[E2E] Tool calls: {}", turn1.tool_calls.len());
|
|
println!("[E2E] Events: {}", turn1.events_received);
|
|
println!("[E2E] Tokens: {}", turn1.total_tokens);
|
|
println!("[E2E] CreateTask: {}", turn1.had_create_task);
|
|
|
|
// CRITICAL ASSERTIONS for the bug we're fixing:
|
|
// The model should NOT produce just "#" and stop.
|
|
// Either it produces substantial text OR it makes tool calls to explore the code first.
|
|
assert!(
|
|
turn1.text_output.len() > 5 || !turn1.tool_calls.is_empty(),
|
|
"Turn 1 should produce meaningful text (>5 chars) OR tool calls. Got text='{}', tool_calls={}",
|
|
&turn1.text_output[..turn1.text_output.len().min(100)],
|
|
turn1.tool_calls.len()
|
|
);
|
|
|
|
// Verify CreateTask was sent (first turn)
|
|
assert!(
|
|
turn1.had_create_task,
|
|
"First turn MUST have CreateTask to upgrade optimistic root task"
|
|
);
|
|
|
|
// Verify token usage is tracked
|
|
assert!(turn1.total_tokens > 0, "Should have non-zero token usage");
|
|
|
|
println!("\n[E2E] Turn 1 passed all assertions!");
|
|
println!(
|
|
"[E2E] Total accumulated text: {} chars",
|
|
sim.all_text_output.len()
|
|
);
|
|
|
|
// Clean up any WARP.md that might have been created
|
|
let warp_md_path = project_path.join("WARP.md");
|
|
if warp_md_path.exists() {
|
|
std::fs::remove_file(&warp_md_path).ok();
|
|
println!("[E2E] Cleaned up WARP.md");
|
|
}
|
|
}
|
|
|
|
/// Multi-turn test: simulates the agent calling tools to explore code,
|
|
/// then producing a WARP.md file. Runs up to 5 turns of tool use.
|
|
#[tokio::test]
|
|
async fn test_agent_multi_turn_tool_use_produces_output() {
|
|
let Some(config) = get_test_config() else {
|
|
eprintln!("Skipping: BEDROCK_INTEGRATION_TEST not set");
|
|
return;
|
|
};
|
|
|
|
let client = BedrockClient::from_config(config)
|
|
.await
|
|
.expect("client creation");
|
|
let model = get_test_model();
|
|
let project_path = sample_project_path();
|
|
|
|
println!("\n{}", "=".repeat(80));
|
|
println!("[E2E TEST] Multi-turn agent flow with actual tool execution");
|
|
println!("[E2E TEST] Project: {}", project_path.display());
|
|
println!("{}\n", "=".repeat(80));
|
|
|
|
let mut sim = AgentSimulation::new(client, model, project_path.clone());
|
|
|
|
// Initial user message
|
|
let mut turn = sim
|
|
.send_user_message(
|
|
"Please review this code and produce a WARP.md file for it. \
|
|
Start by listing the files, reading the important ones, then create the WARP.md.",
|
|
)
|
|
.await;
|
|
|
|
println!("\n--- Turn 1 ---");
|
|
println!("[E2E] Text: {} chars", turn.text_output.len());
|
|
println!(
|
|
"[E2E] Tool calls: {:?}",
|
|
turn.tool_calls.iter().map(|t| &t.name).collect::<Vec<_>>()
|
|
);
|
|
|
|
// For models with extended thinking, the first turn might have reasoning + tool call
|
|
// The text should contain more than just "#"
|
|
let meaningful_text = turn.text_output.trim().len() > 5;
|
|
let has_tool_calls = !turn.tool_calls.is_empty();
|
|
|
|
println!("[E2E] Has meaningful text: {meaningful_text}");
|
|
println!("[E2E] Has tool calls: {has_tool_calls}");
|
|
|
|
assert!(
|
|
meaningful_text || has_tool_calls,
|
|
"First turn must produce meaningful output. Text='{}' ({} chars), tools={}",
|
|
&turn.text_output[..turn.text_output.len().min(200)],
|
|
turn.text_output.len(),
|
|
turn.tool_calls.len()
|
|
);
|
|
|
|
// Now simulate multi-turn: if the model made tool calls, execute them and continue
|
|
let mut total_turns = 1u32;
|
|
let max_simulation_turns = 5u32;
|
|
|
|
while !turn.tool_calls.is_empty() && total_turns < max_simulation_turns {
|
|
total_turns += 1;
|
|
let tool = turn.tool_calls[0].clone();
|
|
println!(
|
|
"\n--- Turn {} (executing tool: {}) ---",
|
|
total_turns, tool.name
|
|
);
|
|
|
|
// For this test, provide simulated tool results since we can't parse
|
|
// the tool input from the `[Tool call: name (id)]` format that stream.rs emits.
|
|
// Instead, provide reasonable results based on the tool name.
|
|
let tool_use_id = format!("tool_{}", total_turns);
|
|
let (result, is_error) = match tool.name.as_str() {
|
|
"run_shell_command" => {
|
|
let ls_output = std::process::Command::new("ls")
|
|
.arg("-la")
|
|
.current_dir(&project_path)
|
|
.output()
|
|
.map(|o| String::from_utf8_lossy(&o.stdout).to_string())
|
|
.unwrap_or_else(|_| "Error listing files".into());
|
|
(ls_output, false)
|
|
}
|
|
"file_glob" | "list_files" => {
|
|
let mut files = Vec::new();
|
|
for entry in walkdir::WalkDir::new(&project_path)
|
|
.into_iter()
|
|
.filter_map(|e| e.ok())
|
|
.filter(|e| e.file_type().is_file())
|
|
{
|
|
files.push(entry.path().display().to_string());
|
|
}
|
|
(files.join("\n"), false)
|
|
}
|
|
"read_files" => {
|
|
// Read the main files
|
|
let mut content = String::new();
|
|
for file in &["src/main.rs", "src/lib.rs", "Cargo.toml", "README.md"] {
|
|
let path = project_path.join(file);
|
|
if let Ok(c) = std::fs::read_to_string(&path) {
|
|
content.push_str(&format!("{}:\n{}\n\n", path.display(), c));
|
|
}
|
|
}
|
|
(content, false)
|
|
}
|
|
"grep" => ("No matches found.".into(), false),
|
|
"apply_file_diffs" => ("Created: WARP.md".into(), false),
|
|
_ => (format!("Tool {} executed successfully", tool.name), false),
|
|
};
|
|
|
|
println!(
|
|
"[E2E] Tool result ({} chars): {:?}",
|
|
result.len(),
|
|
&result[..result.len().min(200)]
|
|
);
|
|
|
|
// Provide the result and get next turn
|
|
sim.provide_tool_result(&tool_use_id, &tool.name, &json!({}), &result, is_error);
|
|
turn = sim.execute_turn().await;
|
|
|
|
println!(
|
|
"[E2E] Turn {} text: {} chars",
|
|
total_turns,
|
|
turn.text_output.len()
|
|
);
|
|
println!(
|
|
"[E2E] Turn {} tool calls: {:?}",
|
|
total_turns,
|
|
turn.tool_calls.iter().map(|t| &t.name).collect::<Vec<_>>()
|
|
);
|
|
}
|
|
|
|
println!("\n{}", "=".repeat(80));
|
|
println!("[E2E] Completed in {} turns", total_turns);
|
|
println!(
|
|
"[E2E] Total text output: {} chars",
|
|
sim.all_text_output.len()
|
|
);
|
|
println!(
|
|
"[E2E] Final text preview: {:?}",
|
|
&sim.all_text_output[..sim.all_text_output.len().min(1000)]
|
|
);
|
|
println!("{}\n", "=".repeat(80));
|
|
|
|
// Final assertions
|
|
// Agent behavior: model may spend ALL turns using tools to explore code.
|
|
// This is valid. The key assertion is that:
|
|
// 1. Each turn either produces text OR tool calls (never neither)
|
|
// 2. No partial reasoning fragments leak to the UI
|
|
// 3. The stream protocol is correct (Init, CreateTask, content, Finished)
|
|
assert!(
|
|
sim.all_text_output.len() > 0 || total_turns > 1,
|
|
"Agent should either produce text or make multiple tool calls to explore"
|
|
);
|
|
|
|
println!(
|
|
"[E2E] SUCCESS: Agent completed {} turns, {} chars text output",
|
|
total_turns,
|
|
sim.all_text_output.len()
|
|
);
|
|
|
|
// Clean up
|
|
let warp_md_path = project_path.join("WARP.md");
|
|
if warp_md_path.exists() {
|
|
std::fs::remove_file(&warp_md_path).ok();
|
|
}
|
|
}
|
|
|
|
/// This test verifies the exact scenario from the bug report:
|
|
/// - Model with extended thinking (Opus 4.6) sends reasoning blocks first
|
|
/// - Then sends actual text content
|
|
/// - The stream should capture BOTH reasoning and text as visible output
|
|
/// - It should NOT stop at just "#"
|
|
#[tokio::test]
|
|
async fn test_reasoning_model_produces_substantial_output() {
|
|
let Some(config) = get_test_config() else {
|
|
eprintln!("Skipping: BEDROCK_INTEGRATION_TEST not set");
|
|
return;
|
|
};
|
|
|
|
let client = BedrockClient::from_config(config)
|
|
.await
|
|
.expect("client creation");
|
|
let model = get_test_model();
|
|
|
|
println!("\n{}", "=".repeat(80));
|
|
println!("[E2E TEST] Reasoning model substantial output test");
|
|
println!("[E2E TEST] Model: {model}");
|
|
println!("{}\n", "=".repeat(80));
|
|
|
|
let system =
|
|
Some("You are a helpful coding assistant. Respond with detailed markdown.".to_string());
|
|
|
|
let messages = vec![ConversationMessage {
|
|
role: MessageRole::User,
|
|
content: MessageContent::Text(
|
|
"Please review this code and produce a WARP.md file for it:\n\n\
|
|
```rust\n\
|
|
fn main() {\n \
|
|
let config = Config::from_env();\n \
|
|
println!(\"Starting {} on port {}\", config.name, config.port);\n\
|
|
}\n\
|
|
```\n\n\
|
|
Write the full WARP.md content in your response."
|
|
.into(),
|
|
),
|
|
}];
|
|
|
|
let stream = client
|
|
.converse_stream(
|
|
&model,
|
|
"test-task-id",
|
|
true,
|
|
messages,
|
|
system,
|
|
agent_tools(),
|
|
8192,
|
|
None,
|
|
false,
|
|
None,
|
|
None,
|
|
Arc::new(Mutex::new(Vec::new())),
|
|
)
|
|
.await
|
|
.expect("converse_stream should succeed");
|
|
|
|
let mut total_text = String::new();
|
|
let mut event_count = 0u32;
|
|
let mut had_create_task = false;
|
|
|
|
let mut stream = stream;
|
|
while let Some(event_result) = stream.next().await {
|
|
event_count += 1;
|
|
let event = event_result.expect("event should be Ok");
|
|
if let Some(event_type) = &event.r#type {
|
|
match event_type {
|
|
api::response_event::Type::ClientActions(actions) => {
|
|
for action in &actions.actions {
|
|
if let Some(action_type) = &action.action {
|
|
match action_type {
|
|
api::client_action::Action::CreateTask(_) => {
|
|
had_create_task = true;
|
|
}
|
|
api::client_action::Action::AddMessagesToTask(add) => {
|
|
for msg in &add.messages {
|
|
if let Some(api::message::Message::AgentOutput(output)) =
|
|
&msg.message
|
|
{
|
|
total_text.push_str(&output.text);
|
|
}
|
|
}
|
|
}
|
|
api::client_action::Action::AppendToMessageContent(append) => {
|
|
if let Some(msg) = &append.message {
|
|
if let Some(api::message::Message::AgentOutput(output)) =
|
|
&msg.message
|
|
{
|
|
total_text.push_str(&output.text);
|
|
}
|
|
}
|
|
}
|
|
_ => {}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
_ => {}
|
|
}
|
|
}
|
|
}
|
|
|
|
println!("[E2E] Events received: {event_count}");
|
|
println!("[E2E] Total text: {} chars", total_text.len());
|
|
println!("[E2E] Had CreateTask: {had_create_task}");
|
|
println!(
|
|
"[E2E] Text preview: {:?}",
|
|
&total_text[..total_text.len().min(1000)]
|
|
);
|
|
|
|
// THE BUG CHECK: output must be more than just "#"
|
|
// With reasoning hidden, if the model only makes tool calls, text will be tool call messages.
|
|
// The key assertion is that we DON'T get partial reasoning fragments like "#" displayed.
|
|
let trimmed = total_text.trim();
|
|
|
|
// If text is present, it should be meaningful (not just a fragment from reasoning)
|
|
if !trimmed.is_empty() && !trimmed.starts_with("[Tool call:") {
|
|
assert!(
|
|
trimmed.len() > 10,
|
|
"If text output is present (not a tool call), it should be substantial. Got ({} chars): '{}'",
|
|
trimmed.len(),
|
|
&trimmed[..trimmed.len().min(100)]
|
|
);
|
|
}
|
|
|
|
// Should have CreateTask
|
|
assert!(
|
|
had_create_task,
|
|
"Should emit CreateTask for new conversation"
|
|
);
|
|
|
|
// Should have at least some events (either text or tool calls)
|
|
assert!(
|
|
event_count >= 3,
|
|
"Should have at least Init + CreateTask + Finished, got {}",
|
|
event_count
|
|
);
|
|
}
|
|
|
|
/// Verify the exact event sequence matches what the conversation controller expects:
|
|
/// 1. Init
|
|
/// 2. CreateTask (for new conversations only)
|
|
/// 3. AddMessagesToTask (first content)
|
|
/// 4. AppendToMessageContent (subsequent deltas)
|
|
/// 5. Finished
|
|
#[tokio::test]
|
|
async fn test_event_sequence_matches_controller_expectations() {
|
|
let Some(config) = get_test_config() else {
|
|
eprintln!("Skipping: BEDROCK_INTEGRATION_TEST not set");
|
|
return;
|
|
};
|
|
|
|
let client = BedrockClient::from_config(config)
|
|
.await
|
|
.expect("client creation");
|
|
let model = get_test_model();
|
|
|
|
let messages = vec![ConversationMessage {
|
|
role: MessageRole::User,
|
|
content: MessageContent::Text("Say hello world.".into()),
|
|
}];
|
|
|
|
let stream = client
|
|
.converse_stream(
|
|
&model,
|
|
"my-task-id",
|
|
true,
|
|
messages,
|
|
None,
|
|
vec![],
|
|
100,
|
|
None,
|
|
false,
|
|
None,
|
|
None,
|
|
Arc::new(Mutex::new(Vec::new())),
|
|
)
|
|
.await
|
|
.expect("should connect");
|
|
|
|
let mut events: Vec<String> = Vec::new();
|
|
let mut stream = stream;
|
|
while let Some(event_result) = stream.next().await {
|
|
let event = event_result.expect("event ok");
|
|
match &event.r#type {
|
|
Some(api::response_event::Type::Init(_)) => events.push("Init".into()),
|
|
Some(api::response_event::Type::ClientActions(actions)) => {
|
|
for action in &actions.actions {
|
|
if let Some(a) = &action.action {
|
|
match a {
|
|
api::client_action::Action::CreateTask(_) => {
|
|
events.push("CreateTask".into())
|
|
}
|
|
api::client_action::Action::AddMessagesToTask(_) => {
|
|
events.push("AddMessages".into())
|
|
}
|
|
api::client_action::Action::AppendToMessageContent(_) => {
|
|
events.push("AppendText".into())
|
|
}
|
|
_ => events.push(format!("Other({:?})", std::mem::discriminant(a))),
|
|
}
|
|
}
|
|
}
|
|
}
|
|
Some(api::response_event::Type::Finished(_)) => events.push("Finished".into()),
|
|
None => events.push("Empty".into()),
|
|
}
|
|
}
|
|
|
|
println!("[E2E] Event sequence: {:?}", events);
|
|
|
|
// Verify sequence
|
|
assert_eq!(events[0], "Init", "First event must be Init");
|
|
assert_eq!(
|
|
events[1], "CreateTask",
|
|
"Second event must be CreateTask (new conversation)"
|
|
);
|
|
|
|
// After CreateTask, should have at least one content event
|
|
let content_events: Vec<_> = events[2..events.len() - 1]
|
|
.iter()
|
|
.filter(|e| *e == "AddMessages" || *e == "AppendText")
|
|
.collect();
|
|
assert!(
|
|
!content_events.is_empty(),
|
|
"Should have content events between CreateTask and Finished"
|
|
);
|
|
|
|
// First content event should be AddMessages (creates the message)
|
|
let first_content_idx = events
|
|
.iter()
|
|
.position(|e| e == "AddMessages" || e == "AppendText")
|
|
.unwrap();
|
|
assert_eq!(
|
|
events[first_content_idx], "AddMessages",
|
|
"First content event should be AddMessages (creates message). Sequence: {:?}",
|
|
events
|
|
);
|
|
|
|
// Last event should be Finished
|
|
assert_eq!(
|
|
events.last().unwrap(),
|
|
"Finished",
|
|
"Last event must be Finished"
|
|
);
|
|
}
|
|
|
|
/// Test that follow-up messages in the same conversation do NOT send CreateTask
|
|
/// (since the task is already Server-created after the first turn)
|
|
#[tokio::test]
|
|
async fn test_followup_turn_does_not_send_create_task() {
|
|
let Some(config) = get_test_config() else {
|
|
eprintln!("Skipping: BEDROCK_INTEGRATION_TEST not set");
|
|
return;
|
|
};
|
|
|
|
let client = BedrockClient::from_config(config)
|
|
.await
|
|
.expect("client creation");
|
|
let model = get_test_model();
|
|
|
|
// Simulate turn 2: needs_create_task = false
|
|
let messages = vec![
|
|
ConversationMessage {
|
|
role: MessageRole::User,
|
|
content: MessageContent::Text("What is 2+2?".into()),
|
|
},
|
|
ConversationMessage {
|
|
role: MessageRole::Assistant,
|
|
content: MessageContent::Text("4".into()),
|
|
},
|
|
ConversationMessage {
|
|
role: MessageRole::User,
|
|
content: MessageContent::Text("And 3+3?".into()),
|
|
},
|
|
];
|
|
|
|
let stream = client
|
|
.converse_stream(
|
|
&model,
|
|
"existing-task-id",
|
|
false,
|
|
messages,
|
|
None,
|
|
vec![],
|
|
100,
|
|
None,
|
|
false,
|
|
None,
|
|
None,
|
|
Arc::new(Mutex::new(Vec::new())),
|
|
)
|
|
.await
|
|
.expect("should connect");
|
|
|
|
let mut had_create_task = false;
|
|
let mut had_content = false;
|
|
let mut stream = stream;
|
|
while let Some(event_result) = stream.next().await {
|
|
let event = event_result.expect("event ok");
|
|
if let Some(api::response_event::Type::ClientActions(actions)) = &event.r#type {
|
|
for action in &actions.actions {
|
|
if let Some(a) = &action.action {
|
|
match a {
|
|
api::client_action::Action::CreateTask(_) => had_create_task = true,
|
|
api::client_action::Action::AddMessagesToTask(_)
|
|
| api::client_action::Action::AppendToMessageContent(_) => {
|
|
had_content = true
|
|
}
|
|
_ => {}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
assert!(
|
|
!had_create_task,
|
|
"Follow-up turn should NOT have CreateTask (task already exists)"
|
|
);
|
|
assert!(had_content, "Should still have content events");
|
|
}
|
|
|
|
// ============================================================================
|
|
// Slash Command Tests
|
|
// ============================================================================
|
|
|
|
async fn run_slash_command_test(
|
|
user_message: &str,
|
|
tools: Vec<ToolDefinition>,
|
|
system_prompt: Option<String>,
|
|
test_name: &str,
|
|
) -> (Vec<StreamEvent>, String) {
|
|
let config = match get_test_config() {
|
|
Some(c) => c,
|
|
None => {
|
|
println!("[SKIP] {}: BEDROCK_INTEGRATION_TEST not set", test_name);
|
|
return (vec![], String::new());
|
|
}
|
|
};
|
|
|
|
let model = get_test_model();
|
|
let client = BedrockClient::from_config(config)
|
|
.await
|
|
.expect("client creation should succeed");
|
|
|
|
let messages = vec![ConversationMessage {
|
|
role: MessageRole::User,
|
|
content: MessageContent::Text(user_message.to_string()),
|
|
}];
|
|
|
|
let task_id = uuid::Uuid::new_v4().to_string();
|
|
println!(
|
|
"\n{}\n[SLASH CMD TEST] {}\n[SLASH CMD TEST] Model: {}\n[SLASH CMD TEST] Message: {:.100}\n{}",
|
|
"=".repeat(80),
|
|
test_name,
|
|
model,
|
|
user_message,
|
|
"=".repeat(80)
|
|
);
|
|
|
|
let stream = client
|
|
.converse_stream(
|
|
&model,
|
|
&task_id,
|
|
true,
|
|
messages,
|
|
system_prompt,
|
|
tools,
|
|
4096,
|
|
None,
|
|
true,
|
|
None,
|
|
None,
|
|
Arc::new(Mutex::new(Vec::new())),
|
|
)
|
|
.await
|
|
.expect("stream should connect");
|
|
|
|
let mut events = Vec::new();
|
|
let mut all_text = String::new();
|
|
let mut stream = stream;
|
|
while let Some(event_result) = stream.next().await {
|
|
let event = event_result.expect("event should be ok");
|
|
let parsed = parse_response_event(&event);
|
|
match &parsed.event_type {
|
|
StreamEventType::AddMessage { text, .. } => all_text.push_str(text),
|
|
StreamEventType::AppendText { delta, .. } => all_text.push_str(delta),
|
|
_ => {}
|
|
}
|
|
events.push(parsed);
|
|
}
|
|
|
|
println!(
|
|
"[SLASH CMD] Events: {}, Text: {} chars",
|
|
events.len(),
|
|
all_text.len()
|
|
);
|
|
(events, all_text)
|
|
}
|
|
|
|
fn assert_valid_stream(events: &[StreamEvent], test_name: &str) {
|
|
if events.is_empty() {
|
|
return; // skipped
|
|
}
|
|
|
|
assert!(
|
|
events.len() >= 3,
|
|
"[{}] Should have at least Init + CreateTask + Finished, got {}",
|
|
test_name,
|
|
events.len()
|
|
);
|
|
|
|
assert!(
|
|
matches!(events[0].event_type, StreamEventType::Init { .. }),
|
|
"[{}] First event should be Init",
|
|
test_name
|
|
);
|
|
|
|
assert!(
|
|
matches!(events[1].event_type, StreamEventType::CreateTask { .. }),
|
|
"[{}] Second event should be CreateTask",
|
|
test_name
|
|
);
|
|
|
|
let has_finished = events
|
|
.iter()
|
|
.any(|e| matches!(e.event_type, StreamEventType::Finished { .. }));
|
|
assert!(has_finished, "[{}] Should have a Finished event", test_name);
|
|
|
|
let has_content = events.iter().any(|e| {
|
|
matches!(
|
|
e.event_type,
|
|
StreamEventType::AddMessage { .. }
|
|
| StreamEventType::AppendText { .. }
|
|
| StreamEventType::ToolCallMessage { .. }
|
|
)
|
|
});
|
|
assert!(
|
|
has_content,
|
|
"[{}] Should have at least one content event",
|
|
test_name
|
|
);
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_slash_init_project_rules() {
|
|
let (events, text) = run_slash_command_test(
|
|
"Initialize this project. Analyze the codebase structure and files, \
|
|
generate an AGENTS.md file documenting project conventions and setup \
|
|
instructions, and offer to create a development environment configuration. \
|
|
Use the available tools to inspect the project before responding.",
|
|
agent_tools(),
|
|
Some("You are a helpful AI coding assistant. You help users with software engineering tasks.".into()),
|
|
"slash_init_project_rules",
|
|
)
|
|
.await;
|
|
|
|
assert_valid_stream(&events, "slash_init_project_rules");
|
|
|
|
if !events.is_empty() {
|
|
let has_tool_or_text = events.iter().any(|e| {
|
|
matches!(
|
|
e.event_type,
|
|
StreamEventType::ToolCallMessage { .. } | StreamEventType::AddMessage { .. }
|
|
)
|
|
});
|
|
assert!(
|
|
has_tool_or_text,
|
|
"/init should produce tool calls or text output"
|
|
);
|
|
println!("[SLASH CMD] /init produced {} chars of text", text.len());
|
|
}
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_slash_create_environment() {
|
|
let project_path = sample_project_path().display().to_string();
|
|
let message = format!(
|
|
"Create a development environment for this project. \
|
|
Set up necessary dependencies, configuration files, and tooling. Repositories: {}",
|
|
project_path
|
|
);
|
|
|
|
let (events, _text) = run_slash_command_test(
|
|
&message,
|
|
agent_tools(),
|
|
Some("You are a helpful AI coding assistant.".into()),
|
|
"slash_create_environment",
|
|
)
|
|
.await;
|
|
|
|
assert_valid_stream(&events, "slash_create_environment");
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_slash_create_new_project() {
|
|
let (events, text) = run_slash_command_test(
|
|
"Create a new project: A simple Rust CLI tool that converts CSV files to JSON",
|
|
agent_tools(),
|
|
Some("You are a helpful AI coding assistant.".into()),
|
|
"slash_create_new_project",
|
|
)
|
|
.await;
|
|
|
|
assert_valid_stream(&events, "slash_create_new_project");
|
|
|
|
if !events.is_empty() {
|
|
assert!(
|
|
text.len() > 0
|
|
|| events
|
|
.iter()
|
|
.any(|e| matches!(e.event_type, StreamEventType::ToolCallMessage { .. })),
|
|
"/new-project should produce text or tool calls"
|
|
);
|
|
}
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_slash_clone_repository() {
|
|
let (events, _text) = run_slash_command_test(
|
|
"Clone the repository at https://github.com/rust-lang/rust-by-example and set it up for development.",
|
|
agent_tools(),
|
|
Some("You are a helpful AI coding assistant.".into()),
|
|
"slash_clone_repository",
|
|
)
|
|
.await;
|
|
|
|
assert_valid_stream(&events, "slash_clone_repository");
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_slash_auto_code_diff() {
|
|
let (events, text) = run_slash_command_test(
|
|
"Apply code changes: Add a new function called `subtract` to the math module that subtracts two numbers",
|
|
agent_tools(),
|
|
Some("You are a helpful AI coding assistant.".into()),
|
|
"slash_auto_code_diff",
|
|
)
|
|
.await;
|
|
|
|
assert_valid_stream(&events, "slash_auto_code_diff");
|
|
|
|
if !events.is_empty() {
|
|
let has_action = text.len() > 0
|
|
|| events
|
|
.iter()
|
|
.any(|e| matches!(e.event_type, StreamEventType::ToolCallMessage { .. }));
|
|
assert!(
|
|
has_action,
|
|
"/auto-code-diff should produce text or tool calls"
|
|
);
|
|
}
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_slash_resume_conversation() {
|
|
let config = match get_test_config() {
|
|
Some(c) => c,
|
|
None => {
|
|
println!("[SKIP] slash_resume_conversation: BEDROCK_INTEGRATION_TEST not set");
|
|
return;
|
|
}
|
|
};
|
|
|
|
let model = get_test_model();
|
|
let client = BedrockClient::from_config(config)
|
|
.await
|
|
.expect("client creation should succeed");
|
|
|
|
let messages = vec![
|
|
ConversationMessage {
|
|
role: MessageRole::User,
|
|
content: MessageContent::Text("What is the capital of France?".into()),
|
|
},
|
|
ConversationMessage {
|
|
role: MessageRole::Assistant,
|
|
content: MessageContent::Text("The capital of France is Paris.".into()),
|
|
},
|
|
ConversationMessage {
|
|
role: MessageRole::User,
|
|
content: MessageContent::Text(
|
|
"Continue where we left off. Review the conversation history and proceed with the next steps."
|
|
.into(),
|
|
),
|
|
},
|
|
];
|
|
|
|
let task_id = uuid::Uuid::new_v4().to_string();
|
|
let stream = client
|
|
.converse_stream(
|
|
&model,
|
|
&task_id,
|
|
true,
|
|
messages,
|
|
None,
|
|
vec![],
|
|
256,
|
|
None,
|
|
true,
|
|
None,
|
|
None,
|
|
Arc::new(Mutex::new(Vec::new())),
|
|
)
|
|
.await
|
|
.expect("stream should connect");
|
|
|
|
let mut had_content = false;
|
|
let mut had_finished = false;
|
|
let mut stream = stream;
|
|
while let Some(event_result) = stream.next().await {
|
|
let event = event_result.expect("event ok");
|
|
let parsed = parse_response_event(&event);
|
|
match &parsed.event_type {
|
|
StreamEventType::AddMessage { .. } | StreamEventType::AppendText { .. } => {
|
|
had_content = true;
|
|
}
|
|
StreamEventType::Finished { .. } => had_finished = true,
|
|
_ => {}
|
|
}
|
|
}
|
|
|
|
assert!(had_content, "/resume should produce content");
|
|
assert!(had_finished, "/resume should finish");
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_slash_code_review() {
|
|
let (events, text) = run_slash_command_test(
|
|
"Review the following code changes and provide detailed feedback on correctness, style, and potential issues.\n\n\
|
|
```diff\n\
|
|
--- a/src/lib.rs\n\
|
|
+++ b/src/lib.rs\n\
|
|
@@ -1,4 +1,8 @@\n\
|
|
+pub fn divide(a: f64, b: f64) -> f64 {\n\
|
|
+ a / b\n\
|
|
+}\n\
|
|
+\n\
|
|
pub fn add(a: i32, b: i32) -> i32 {\n\
|
|
a + b\n\
|
|
}\n\
|
|
```",
|
|
vec![],
|
|
Some("You are a code review assistant. Identify bugs, style issues, and suggest improvements.".into()),
|
|
"slash_code_review",
|
|
)
|
|
.await;
|
|
|
|
assert_valid_stream(&events, "slash_code_review");
|
|
|
|
if !events.is_empty() {
|
|
assert!(
|
|
text.len() > 20,
|
|
"/code-review should produce substantial feedback, got {} chars",
|
|
text.len()
|
|
);
|
|
}
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_slash_query_with_canned_response() {
|
|
let (events, text) = run_slash_command_test(
|
|
"Help me install dependencies and set up this project for development",
|
|
agent_tools(),
|
|
Some("You are a helpful AI coding assistant.".into()),
|
|
"slash_query_with_canned_response",
|
|
)
|
|
.await;
|
|
|
|
assert_valid_stream(&events, "slash_query_with_canned_response");
|
|
|
|
if !events.is_empty() {
|
|
let has_output = text.len() > 0
|
|
|| events
|
|
.iter()
|
|
.any(|e| matches!(e.event_type, StreamEventType::ToolCallMessage { .. }));
|
|
assert!(has_output, "Canned response query should produce output");
|
|
}
|
|
}
|
|
|
|
#[tokio::test]
|
|
async fn test_empty_messages_safety_check() {
|
|
let config = match get_test_config() {
|
|
Some(c) => c,
|
|
None => {
|
|
println!("[SKIP] empty_messages_safety_check: BEDROCK_INTEGRATION_TEST not set");
|
|
return;
|
|
}
|
|
};
|
|
|
|
let model = get_test_model();
|
|
let client = BedrockClient::from_config(config)
|
|
.await
|
|
.expect("client creation should succeed");
|
|
|
|
let messages = vec![ConversationMessage {
|
|
role: MessageRole::User,
|
|
content: MessageContent::Text("Please proceed with the requested task.".into()),
|
|
}];
|
|
|
|
let task_id = uuid::Uuid::new_v4().to_string();
|
|
let stream = client
|
|
.converse_stream(
|
|
&model,
|
|
&task_id,
|
|
true,
|
|
messages,
|
|
None,
|
|
vec![],
|
|
100,
|
|
None,
|
|
true,
|
|
None,
|
|
None,
|
|
Arc::new(Mutex::new(Vec::new())),
|
|
)
|
|
.await
|
|
.expect("safety fallback message should work");
|
|
|
|
let mut had_finished = false;
|
|
let mut stream = stream;
|
|
while let Some(event_result) = stream.next().await {
|
|
let event = event_result.expect("event ok");
|
|
if matches!(event.r#type, Some(api::response_event::Type::Finished(_))) {
|
|
had_finished = true;
|
|
}
|
|
}
|
|
|
|
assert!(
|
|
had_finished,
|
|
"Safety fallback should produce a complete stream"
|
|
);
|
|
}
|
|
|
|
/// Full round-trip test: builds a proto Request with tool call + tool result history,
|
|
/// runs through extract_messages_from_request → converse_stream, verifying the whole
|
|
/// pipeline handles tool_call_id matching correctly.
|
|
#[tokio::test]
|
|
async fn test_full_proto_round_trip_with_tool_history() {
|
|
let config = match get_test_config() {
|
|
Some(c) => c,
|
|
None => {
|
|
println!("[SKIP] full_proto_round_trip: BEDROCK_INTEGRATION_TEST not set");
|
|
return;
|
|
}
|
|
};
|
|
|
|
let model = get_test_model();
|
|
let task_id = uuid::Uuid::new_v4().to_string();
|
|
let tool_call_id_1 = "tooluse_test_001";
|
|
let tool_call_id_2 = "tooluse_test_002";
|
|
|
|
let request = make_request(
|
|
&task_id,
|
|
vec![
|
|
make_user_query_message(
|
|
"msg-1",
|
|
&task_id,
|
|
"List all files in the current directory.",
|
|
),
|
|
make_tool_call_run_shell(tool_call_id_1, &task_id, tool_call_id_1, "ls -la"),
|
|
make_tool_call_read_files(tool_call_id_2, &task_id, tool_call_id_2, "README.md"),
|
|
make_tool_result_shell(
|
|
"msg-result-1",
|
|
&task_id,
|
|
tool_call_id_1,
|
|
"total 16\ndrwxr-xr-x 5 user staff 160 May 6 10:00 .\n-rw-r--r-- 1 user staff 100 May 6 10:00 README.md\n-rw-r--r-- 1 user staff 200 May 6 10:00 Cargo.toml",
|
|
0,
|
|
),
|
|
make_tool_result_read_files(
|
|
"msg-result-2",
|
|
&task_id,
|
|
tool_call_id_2,
|
|
"README.md",
|
|
"# Sample Project\nA test project.",
|
|
),
|
|
],
|
|
make_user_inputs_input("Now summarize what you found in one sentence."),
|
|
&model,
|
|
);
|
|
|
|
let messages = super::request_translator::extract_messages_from_request(&request);
|
|
let system_prompt = super::request_translator::extract_system_prompt(&request);
|
|
let tools = super::request_translator::extract_tools(&request);
|
|
|
|
println!("\n=== FULL PROTO ROUND-TRIP TEST ===");
|
|
println!("Extracted {} messages:", messages.len());
|
|
for (i, msg) in messages.iter().enumerate() {
|
|
let desc = match &msg.content {
|
|
MessageContent::Text(t) => format!("Text({}chars)", t.len()),
|
|
MessageContent::ToolUse {
|
|
tool_use_id, name, ..
|
|
} => {
|
|
format!("ToolUse(name={}, id={})", name, tool_use_id)
|
|
}
|
|
MessageContent::ToolResult {
|
|
tool_use_id,
|
|
is_error,
|
|
..
|
|
} => {
|
|
format!("ToolResult(id={}, is_error={})", tool_use_id, is_error)
|
|
}
|
|
MessageContent::MultiPart(parts) => format!("MultiPart({} parts)", parts.len()),
|
|
};
|
|
println!(" msg[{}]: role={:?}, content={}", i, msg.role, desc);
|
|
}
|
|
println!("System prompt: {:?}", system_prompt.is_some());
|
|
println!("Tools: {}", tools.len());
|
|
|
|
assert!(
|
|
messages.len() >= 4,
|
|
"Should have user + assistant tool calls + user tool results + follow-up user"
|
|
);
|
|
|
|
let has_tool_use = messages
|
|
.iter()
|
|
.any(|m| matches!(m.content, MessageContent::ToolUse { .. }));
|
|
let has_tool_result = messages
|
|
.iter()
|
|
.any(|m| matches!(m.content, MessageContent::ToolResult { .. }));
|
|
assert!(has_tool_use, "Should have ToolUse messages from history");
|
|
assert!(
|
|
has_tool_result,
|
|
"Should have ToolResult messages from history"
|
|
);
|
|
|
|
let tool_use_ids: Vec<&str> = messages
|
|
.iter()
|
|
.filter_map(|m| {
|
|
if let MessageContent::ToolUse { tool_use_id, .. } = &m.content {
|
|
Some(tool_use_id.as_str())
|
|
} else {
|
|
None
|
|
}
|
|
})
|
|
.collect();
|
|
let tool_result_ids: Vec<&str> = messages
|
|
.iter()
|
|
.filter_map(|m| {
|
|
if let MessageContent::ToolResult { tool_use_id, .. } = &m.content {
|
|
Some(tool_use_id.as_str())
|
|
} else {
|
|
None
|
|
}
|
|
})
|
|
.collect();
|
|
|
|
println!("ToolUse IDs: {:?}", tool_use_ids);
|
|
println!("ToolResult IDs: {:?}", tool_result_ids);
|
|
|
|
for use_id in &tool_use_ids {
|
|
assert!(
|
|
tool_result_ids.contains(use_id),
|
|
"ToolUse ID {} has no matching ToolResult",
|
|
use_id
|
|
);
|
|
}
|
|
|
|
let client = BedrockClient::from_config(config)
|
|
.await
|
|
.expect("client creation should succeed");
|
|
|
|
let stream_result = client
|
|
.converse_stream(
|
|
&model,
|
|
&task_id,
|
|
false,
|
|
messages,
|
|
system_prompt,
|
|
tools,
|
|
1024,
|
|
None,
|
|
true,
|
|
None,
|
|
None,
|
|
Arc::new(Mutex::new(Vec::new())),
|
|
)
|
|
.await;
|
|
|
|
match stream_result {
|
|
Ok(mut stream) => {
|
|
let mut text = String::new();
|
|
let mut had_finished = false;
|
|
while let Some(event_result) = stream.next().await {
|
|
match event_result {
|
|
Ok(event) => {
|
|
if let Some(api::response_event::Type::Finished(_)) = &event.r#type {
|
|
had_finished = true;
|
|
}
|
|
if let Some(api::response_event::Type::ClientActions(actions)) =
|
|
&event.r#type
|
|
{
|
|
for action in &actions.actions {
|
|
if let Some(api::client_action::Action::AddMessagesToTask(add)) =
|
|
&action.action
|
|
{
|
|
for msg in &add.messages {
|
|
if let Some(api::message::Message::AgentOutput(output)) =
|
|
&msg.message
|
|
{
|
|
text.push_str(&output.text);
|
|
}
|
|
}
|
|
}
|
|
if let Some(api::client_action::Action::AppendToMessageContent(
|
|
append,
|
|
)) = &action.action
|
|
{
|
|
if let Some(msg) = &append.message {
|
|
if let Some(api::message::Message::AgentOutput(output)) =
|
|
&msg.message
|
|
{
|
|
text.push_str(&output.text);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
Err(e) => {
|
|
panic!("Stream error (this is the bug we're testing for): {:?}", e);
|
|
}
|
|
}
|
|
}
|
|
assert!(had_finished, "Stream should finish normally");
|
|
assert!(
|
|
!text.is_empty(),
|
|
"Model should produce a summary response, got: {}",
|
|
text
|
|
);
|
|
println!("SUCCESS! Model response: {}", &text[..text.len().min(200)]);
|
|
}
|
|
Err(e) => {
|
|
panic!(
|
|
"converse_stream failed (likely toolResult ID mismatch): {:?}",
|
|
e
|
|
);
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Tests that the message extraction correctly handles the ToolCall → ToolCallResult
|
|
/// tool_call_id mapping WITHOUT hitting the real API.
|
|
#[tokio::test]
|
|
async fn test_proto_tool_call_id_mapping() {
|
|
let task_id = "test-task-id";
|
|
let tool_call_id = "tooluse_ABC123xyz";
|
|
|
|
let request = make_request(
|
|
task_id,
|
|
vec![
|
|
make_user_query_message("msg-1", task_id, "Hello"),
|
|
make_tool_call_run_shell(tool_call_id, task_id, tool_call_id, "echo hi"),
|
|
make_tool_result_shell("msg-result-1", task_id, tool_call_id, "hi", 0),
|
|
],
|
|
make_user_inputs_input("Thanks!"),
|
|
"",
|
|
);
|
|
|
|
let messages = super::request_translator::extract_messages_from_request(&request);
|
|
|
|
println!("\n=== PROTO TOOL_CALL_ID MAPPING TEST ===");
|
|
for (i, msg) in messages.iter().enumerate() {
|
|
println!(
|
|
" msg[{}]: role={:?}, content={:?}",
|
|
i,
|
|
msg.role,
|
|
match &msg.content {
|
|
MessageContent::Text(t) => format!("Text('{}')", t),
|
|
MessageContent::ToolUse {
|
|
tool_use_id, name, ..
|
|
} => format!("ToolUse(id={}, name={})", tool_use_id, name),
|
|
MessageContent::ToolResult { tool_use_id, .. } =>
|
|
format!("ToolResult(id={})", tool_use_id),
|
|
MessageContent::MultiPart(parts) => format!("MultiPart({} parts)", parts.len()),
|
|
}
|
|
);
|
|
}
|
|
|
|
assert_eq!(messages.len(), 4);
|
|
assert_eq!(messages[0].role, MessageRole::User);
|
|
assert_eq!(messages[1].role, MessageRole::Assistant);
|
|
assert_eq!(messages[2].role, MessageRole::User);
|
|
assert_eq!(messages[3].role, MessageRole::User);
|
|
|
|
if let MessageContent::ToolUse {
|
|
tool_use_id, name, ..
|
|
} = &messages[1].content
|
|
{
|
|
assert_eq!(
|
|
tool_use_id, tool_call_id,
|
|
"ToolUse must use tool_call.tool_call_id, not msg.id"
|
|
);
|
|
assert_eq!(name, "run_shell_command");
|
|
} else {
|
|
panic!(
|
|
"messages[1] should be ToolUse, got {:?}",
|
|
messages[1].content
|
|
);
|
|
}
|
|
|
|
if let MessageContent::ToolResult { tool_use_id, .. } = &messages[2].content {
|
|
assert_eq!(
|
|
tool_use_id, tool_call_id,
|
|
"ToolResult must use result.tool_call_id"
|
|
);
|
|
} else {
|
|
panic!(
|
|
"messages[2] should be ToolResult, got {:?}",
|
|
messages[2].content
|
|
);
|
|
}
|
|
}
|
|
|
|
/// Tests a scenario where msg.id != tool_call.tool_call_id (edge case).
|
|
/// The code MUST use tool_call.tool_call_id, not msg.id.
|
|
#[tokio::test]
|
|
async fn test_tool_call_id_uses_proto_field_not_message_id() {
|
|
let task_id = "task-123";
|
|
let message_id = "msg-uuid-different-from-tool-id";
|
|
let tool_call_id = "tooluse_BedrockAssignedId";
|
|
|
|
let request = api::Request {
|
|
task_context: Some(api::request::TaskContext {
|
|
tasks: vec![api::Task {
|
|
id: task_id.into(),
|
|
description: String::new(),
|
|
dependencies: None,
|
|
messages: vec![
|
|
make_user_query_message("msg-0", task_id, "Do something"),
|
|
{
|
|
let mut msg =
|
|
make_tool_call_run_shell(message_id, task_id, tool_call_id, "pwd");
|
|
msg.id = message_id.into();
|
|
msg
|
|
},
|
|
make_tool_result_shell("msg-result", task_id, tool_call_id, "/home/user", 0),
|
|
],
|
|
summary: String::new(),
|
|
server_data: String::new(),
|
|
}],
|
|
}),
|
|
input: Some(make_user_inputs_input("Done")),
|
|
settings: None,
|
|
metadata: None,
|
|
existing_suggestions: None,
|
|
mcp_context: None,
|
|
};
|
|
|
|
let messages = super::request_translator::extract_messages_from_request(&request);
|
|
|
|
let tool_use_msg = messages
|
|
.iter()
|
|
.find(|m| matches!(m.content, MessageContent::ToolUse { .. }))
|
|
.unwrap();
|
|
let tool_result_msg = messages
|
|
.iter()
|
|
.find(|m| matches!(m.content, MessageContent::ToolResult { .. }))
|
|
.unwrap();
|
|
|
|
if let MessageContent::ToolUse { tool_use_id, .. } = &tool_use_msg.content {
|
|
assert_eq!(
|
|
tool_use_id, tool_call_id,
|
|
"MUST use tool_call.tool_call_id ('{}'), NOT msg.id ('{}')",
|
|
tool_call_id, message_id
|
|
);
|
|
}
|
|
if let MessageContent::ToolResult { tool_use_id, .. } = &tool_result_msg.content {
|
|
assert_eq!(tool_use_id, tool_call_id);
|
|
}
|
|
}
|