mirror of
https://github.com/pchuan98/codex.git
synced 2026-07-01 00:31:56 +08:00
feat: dynamic tools injection (#9539)
## Summary Add dynamic tool injection to thread startup in API v2, wire dynamic tool calls through the app server to clients, and plumb responses back into the model tool pipeline. ### Flow (high level) - Thread start injects `dynamic_tools` into the model tool list for that thread (validation is done here). - When the model emits a tool call for one of those names, core raises a `DynamicToolCallRequest` event. - The app server forwards it to the client as `item/tool/call`, waits for the client’s response, then submits a `DynamicToolResponse` back to core. - Core turns that into a `function_call_output` in the next model request so the model can continue. ### What changed - Added dynamic tool specs to v2 thread start params and protocol types; introduced `item/tool/call` (request/response) for dynamic tool execution. - Core now registers dynamic tool specs at request time and routes those calls via a new dynamic tool handler. - App server validates tool names/schemas, forwards dynamic tool call requests to clients, and publishes tool outputs back into the session. - Integration tests
This commit is contained in:
committed by
GitHub
Unverified
parent
25fccc3d4d
commit
d594693d1a
@@ -524,6 +524,12 @@ server_request_definitions! {
|
||||
response: v2::ToolRequestUserInputResponse,
|
||||
},
|
||||
|
||||
/// Execute a dynamic tool call on the client.
|
||||
DynamicToolCall => "item/tool/call" {
|
||||
params: v2::DynamicToolCallParams,
|
||||
response: v2::DynamicToolCallResponse,
|
||||
},
|
||||
|
||||
/// DEPRECATED APIs below
|
||||
/// Request to approve a patch.
|
||||
/// This request is used for Turns started via the legacy APIs (i.e. SendUserTurn, SendUserMessage).
|
||||
|
||||
@@ -325,6 +325,15 @@ pub struct ToolsV2 {
|
||||
pub view_image: Option<bool>,
|
||||
}
|
||||
|
||||
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
|
||||
#[serde(rename_all = "camelCase")]
|
||||
#[ts(export_to = "v2/")]
|
||||
pub struct DynamicToolSpec {
|
||||
pub name: String,
|
||||
pub description: String,
|
||||
pub input_schema: JsonValue,
|
||||
}
|
||||
|
||||
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
|
||||
#[serde(rename_all = "snake_case")]
|
||||
#[ts(export_to = "v2/")]
|
||||
@@ -1090,6 +1099,7 @@ pub struct ThreadStartParams {
|
||||
pub developer_instructions: Option<String>,
|
||||
pub personality: Option<Personality>,
|
||||
pub ephemeral: Option<bool>,
|
||||
pub dynamic_tools: Option<Vec<DynamicToolSpec>>,
|
||||
/// If true, opt into emitting raw response items on the event stream.
|
||||
///
|
||||
/// This is for internal use only (e.g. Codex Cloud).
|
||||
@@ -2374,6 +2384,25 @@ pub struct FileChangeRequestApprovalResponse {
|
||||
pub decision: FileChangeApprovalDecision,
|
||||
}
|
||||
|
||||
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
|
||||
#[serde(rename_all = "camelCase")]
|
||||
#[ts(export_to = "v2/")]
|
||||
pub struct DynamicToolCallParams {
|
||||
pub thread_id: String,
|
||||
pub turn_id: String,
|
||||
pub call_id: String,
|
||||
pub tool: String,
|
||||
pub arguments: JsonValue,
|
||||
}
|
||||
|
||||
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
|
||||
#[serde(rename_all = "camelCase")]
|
||||
#[ts(export_to = "v2/")]
|
||||
pub struct DynamicToolCallResponse {
|
||||
pub output: String,
|
||||
pub success: bool,
|
||||
}
|
||||
|
||||
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
|
||||
#[serde(rename_all = "camelCase")]
|
||||
#[ts(export_to = "v2/")]
|
||||
|
||||
@@ -114,7 +114,20 @@ Start a fresh thread when you need a new Codex conversation.
|
||||
"cwd": "/Users/me/project",
|
||||
"approvalPolicy": "never",
|
||||
"sandbox": "workspaceWrite",
|
||||
"personality": "friendly"
|
||||
"personality": "friendly",
|
||||
"dynamicTools": [
|
||||
{
|
||||
"name": "lookup_ticket",
|
||||
"description": "Fetch a ticket by id",
|
||||
"inputSchema": {
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"id": { "type": "string" }
|
||||
},
|
||||
"required": ["id"]
|
||||
}
|
||||
}
|
||||
],
|
||||
} }
|
||||
{ "id": 10, "result": {
|
||||
"thread": {
|
||||
|
||||
@@ -25,6 +25,7 @@ use codex_app_server_protocol::CommandExecutionRequestApprovalResponse;
|
||||
use codex_app_server_protocol::CommandExecutionStatus;
|
||||
use codex_app_server_protocol::ContextCompactedNotification;
|
||||
use codex_app_server_protocol::DeprecationNoticeNotification;
|
||||
use codex_app_server_protocol::DynamicToolCallParams;
|
||||
use codex_app_server_protocol::ErrorNotification;
|
||||
use codex_app_server_protocol::ExecCommandApprovalParams;
|
||||
use codex_app_server_protocol::ExecCommandApprovalResponse;
|
||||
@@ -85,6 +86,7 @@ use codex_core::protocol::TurnDiffEvent;
|
||||
use codex_core::review_format::format_review_findings_block;
|
||||
use codex_core::review_prompts;
|
||||
use codex_protocol::ThreadId;
|
||||
use codex_protocol::dynamic_tools::DynamicToolResponse as CoreDynamicToolResponse;
|
||||
use codex_protocol::plan_tool::UpdatePlanArgs;
|
||||
use codex_protocol::protocol::ReviewOutputEvent;
|
||||
use codex_protocol::request_user_input::RequestUserInputAnswer as CoreRequestUserInputAnswer;
|
||||
@@ -318,6 +320,40 @@ pub(crate) async fn apply_bespoke_event_handling(
|
||||
}
|
||||
}
|
||||
}
|
||||
EventMsg::DynamicToolCallRequest(request) => {
|
||||
if matches!(api_version, ApiVersion::V2) {
|
||||
let call_id = request.call_id;
|
||||
let params = DynamicToolCallParams {
|
||||
thread_id: conversation_id.to_string(),
|
||||
turn_id: request.turn_id,
|
||||
call_id: call_id.clone(),
|
||||
tool: request.tool,
|
||||
arguments: request.arguments,
|
||||
};
|
||||
let rx = outgoing
|
||||
.send_request(ServerRequestPayload::DynamicToolCall(params))
|
||||
.await;
|
||||
tokio::spawn(async move {
|
||||
crate::dynamic_tools::on_call_response(call_id, rx, conversation).await;
|
||||
});
|
||||
} else {
|
||||
error!(
|
||||
"dynamic tool calls are only supported on api v2 (call_id: {})",
|
||||
request.call_id
|
||||
);
|
||||
let call_id = request.call_id;
|
||||
let _ = conversation
|
||||
.submit(Op::DynamicToolResponse {
|
||||
id: call_id.clone(),
|
||||
response: CoreDynamicToolResponse {
|
||||
call_id,
|
||||
output: "dynamic tool calls require api v2".to_string(),
|
||||
success: false,
|
||||
},
|
||||
})
|
||||
.await;
|
||||
}
|
||||
}
|
||||
// TODO(celia): properly construct McpToolCall TurnItem in core.
|
||||
EventMsg::McpToolCallBegin(begin_event) => {
|
||||
let notification = construct_mcp_tool_call_notification(
|
||||
|
||||
@@ -31,6 +31,7 @@ use codex_app_server_protocol::CollaborationModeListResponse;
|
||||
use codex_app_server_protocol::CommandExecParams;
|
||||
use codex_app_server_protocol::ConversationGitInfo;
|
||||
use codex_app_server_protocol::ConversationSummary;
|
||||
use codex_app_server_protocol::DynamicToolSpec as ApiDynamicToolSpec;
|
||||
use codex_app_server_protocol::ExecOneOffCommandResponse;
|
||||
use codex_app_server_protocol::FeedbackUploadParams;
|
||||
use codex_app_server_protocol::FeedbackUploadResponse;
|
||||
@@ -171,6 +172,7 @@ use codex_login::run_login_server;
|
||||
use codex_protocol::ThreadId;
|
||||
use codex_protocol::config_types::ForcedLoginMethod;
|
||||
use codex_protocol::config_types::Personality;
|
||||
use codex_protocol::dynamic_tools::DynamicToolSpec as CoreDynamicToolSpec;
|
||||
use codex_protocol::items::TurnItem;
|
||||
use codex_protocol::models::ResponseItem;
|
||||
use codex_protocol::protocol::AgentStatus;
|
||||
@@ -1411,35 +1413,81 @@ impl CodexMessageProcessor {
|
||||
}
|
||||
|
||||
async fn thread_start(&mut self, request_id: RequestId, params: ThreadStartParams) {
|
||||
let ThreadStartParams {
|
||||
model,
|
||||
model_provider,
|
||||
cwd,
|
||||
approval_policy,
|
||||
sandbox,
|
||||
config,
|
||||
base_instructions,
|
||||
developer_instructions,
|
||||
dynamic_tools,
|
||||
experimental_raw_events,
|
||||
personality,
|
||||
ephemeral,
|
||||
} = params;
|
||||
let mut typesafe_overrides = self.build_thread_config_overrides(
|
||||
params.model,
|
||||
params.model_provider,
|
||||
params.cwd,
|
||||
params.approval_policy,
|
||||
params.sandbox,
|
||||
params.base_instructions,
|
||||
params.developer_instructions,
|
||||
params.personality,
|
||||
model,
|
||||
model_provider,
|
||||
cwd,
|
||||
approval_policy,
|
||||
sandbox,
|
||||
base_instructions,
|
||||
developer_instructions,
|
||||
personality,
|
||||
);
|
||||
typesafe_overrides.ephemeral = Some(params.ephemeral.unwrap_or_default());
|
||||
typesafe_overrides.ephemeral = ephemeral;
|
||||
|
||||
let config =
|
||||
match derive_config_from_params(&self.cli_overrides, params.config, typesafe_overrides)
|
||||
.await
|
||||
{
|
||||
Ok(config) => config,
|
||||
Err(err) => {
|
||||
let error = JSONRPCErrorError {
|
||||
code: INVALID_REQUEST_ERROR_CODE,
|
||||
message: format!("error deriving config: {err}"),
|
||||
data: None,
|
||||
};
|
||||
self.outgoing.send_error(request_id, error).await;
|
||||
return;
|
||||
}
|
||||
};
|
||||
let config = match derive_config_from_params(
|
||||
&self.cli_overrides,
|
||||
config,
|
||||
typesafe_overrides,
|
||||
)
|
||||
.await
|
||||
{
|
||||
Ok(config) => config,
|
||||
Err(err) => {
|
||||
let error = JSONRPCErrorError {
|
||||
code: INVALID_REQUEST_ERROR_CODE,
|
||||
message: format!("error deriving config: {err}"),
|
||||
data: None,
|
||||
};
|
||||
self.outgoing.send_error(request_id, error).await;
|
||||
return;
|
||||
}
|
||||
};
|
||||
|
||||
match self.thread_manager.start_thread(config).await {
|
||||
let dynamic_tools = dynamic_tools.unwrap_or_default();
|
||||
let core_dynamic_tools = if dynamic_tools.is_empty() {
|
||||
Vec::new()
|
||||
} else {
|
||||
let snapshot = collect_mcp_snapshot(&config).await;
|
||||
let mcp_tool_names = snapshot.tools.keys().cloned().collect::<HashSet<_>>();
|
||||
if let Err(message) = validate_dynamic_tools(&dynamic_tools, &mcp_tool_names) {
|
||||
let error = JSONRPCErrorError {
|
||||
code: INVALID_REQUEST_ERROR_CODE,
|
||||
message,
|
||||
data: None,
|
||||
};
|
||||
self.outgoing.send_error(request_id, error).await;
|
||||
return;
|
||||
}
|
||||
dynamic_tools
|
||||
.into_iter()
|
||||
.map(|tool| CoreDynamicToolSpec {
|
||||
name: tool.name,
|
||||
description: tool.description,
|
||||
input_schema: tool.input_schema,
|
||||
})
|
||||
.collect()
|
||||
};
|
||||
|
||||
match self
|
||||
.thread_manager
|
||||
.start_thread_with_tools(config, core_dynamic_tools)
|
||||
.await
|
||||
{
|
||||
Ok(new_conv) => {
|
||||
let NewThread {
|
||||
thread_id,
|
||||
@@ -1489,7 +1537,7 @@ impl CodexMessageProcessor {
|
||||
if let Err(err) = self
|
||||
.attach_conversation_listener(
|
||||
thread_id,
|
||||
params.experimental_raw_events,
|
||||
experimental_raw_events,
|
||||
ApiVersion::V2,
|
||||
)
|
||||
.await
|
||||
@@ -4322,6 +4370,41 @@ fn errors_to_info(
|
||||
.collect()
|
||||
}
|
||||
|
||||
fn validate_dynamic_tools(
|
||||
tools: &[ApiDynamicToolSpec],
|
||||
mcp_tool_names: &HashSet<String>,
|
||||
) -> Result<(), String> {
|
||||
let mut seen = HashSet::new();
|
||||
for tool in tools {
|
||||
let name = tool.name.trim();
|
||||
if name.is_empty() {
|
||||
return Err("dynamic tool name must not be empty".to_string());
|
||||
}
|
||||
if name != tool.name {
|
||||
return Err(format!(
|
||||
"dynamic tool name has leading/trailing whitespace: {}",
|
||||
tool.name
|
||||
));
|
||||
}
|
||||
if name == "mcp" || name.starts_with("mcp__") {
|
||||
return Err(format!("dynamic tool name is reserved: {name}"));
|
||||
}
|
||||
if mcp_tool_names.contains(name) {
|
||||
return Err(format!("dynamic tool name conflicts with MCP tool: {name}"));
|
||||
}
|
||||
if !seen.insert(name.to_string()) {
|
||||
return Err(format!("duplicate dynamic tool name: {name}"));
|
||||
}
|
||||
|
||||
if let Err(err) = codex_core::parse_tool_input_schema(&tool.input_schema) {
|
||||
return Err(format!(
|
||||
"dynamic tool input schema is not supported for {name}: {err}"
|
||||
));
|
||||
}
|
||||
}
|
||||
Ok(())
|
||||
}
|
||||
|
||||
/// Derive the effective [`Config`] by layering three override sources.
|
||||
///
|
||||
/// Precedence (lowest to highest):
|
||||
@@ -4602,6 +4685,28 @@ mod tests {
|
||||
use serde_json::json;
|
||||
use tempfile::TempDir;
|
||||
|
||||
#[test]
|
||||
fn validate_dynamic_tools_rejects_unsupported_input_schema() {
|
||||
let tools = vec![ApiDynamicToolSpec {
|
||||
name: "my_tool".to_string(),
|
||||
description: "test".to_string(),
|
||||
input_schema: json!({"type": "null"}),
|
||||
}];
|
||||
let err = validate_dynamic_tools(&tools, &HashSet::new()).expect_err("invalid schema");
|
||||
assert!(err.contains("my_tool"), "unexpected error: {err}");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn validate_dynamic_tools_accepts_sanitizable_input_schema() {
|
||||
let tools = vec![ApiDynamicToolSpec {
|
||||
name: "my_tool".to_string(),
|
||||
description: "test".to_string(),
|
||||
// Missing `type` is common; core sanitizes these to a supported schema.
|
||||
input_schema: json!({"properties": {}}),
|
||||
}];
|
||||
validate_dynamic_tools(&tools, &HashSet::new()).expect("valid schema");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn extract_conversation_summary_prefers_plain_user_messages() -> Result<()> {
|
||||
let conversation_id = ThreadId::from_string("3f941c35-29b3-493b-b0a4-e25800d9aeb0")?;
|
||||
|
||||
@@ -0,0 +1,58 @@
|
||||
use codex_app_server_protocol::DynamicToolCallResponse;
|
||||
use codex_core::CodexThread;
|
||||
use codex_protocol::dynamic_tools::DynamicToolResponse as CoreDynamicToolResponse;
|
||||
use codex_protocol::protocol::Op;
|
||||
use std::sync::Arc;
|
||||
use tokio::sync::oneshot;
|
||||
use tracing::error;
|
||||
|
||||
pub(crate) async fn on_call_response(
|
||||
call_id: String,
|
||||
receiver: oneshot::Receiver<serde_json::Value>,
|
||||
conversation: Arc<CodexThread>,
|
||||
) {
|
||||
let response = receiver.await;
|
||||
let value = match response {
|
||||
Ok(value) => value,
|
||||
Err(err) => {
|
||||
error!("request failed: {err:?}");
|
||||
let fallback = CoreDynamicToolResponse {
|
||||
call_id: call_id.clone(),
|
||||
output: "dynamic tool request failed".to_string(),
|
||||
success: false,
|
||||
};
|
||||
if let Err(err) = conversation
|
||||
.submit(Op::DynamicToolResponse {
|
||||
id: call_id.clone(),
|
||||
response: fallback,
|
||||
})
|
||||
.await
|
||||
{
|
||||
error!("failed to submit DynamicToolResponse: {err}");
|
||||
}
|
||||
return;
|
||||
}
|
||||
};
|
||||
|
||||
let response = serde_json::from_value::<DynamicToolCallResponse>(value).unwrap_or_else(|err| {
|
||||
error!("failed to deserialize DynamicToolCallResponse: {err}");
|
||||
DynamicToolCallResponse {
|
||||
output: "dynamic tool response was invalid".to_string(),
|
||||
success: false,
|
||||
}
|
||||
});
|
||||
let response = CoreDynamicToolResponse {
|
||||
call_id: call_id.clone(),
|
||||
output: response.output,
|
||||
success: response.success,
|
||||
};
|
||||
if let Err(err) = conversation
|
||||
.submit(Op::DynamicToolResponse {
|
||||
id: call_id,
|
||||
response,
|
||||
})
|
||||
.await
|
||||
{
|
||||
error!("failed to submit DynamicToolResponse: {err}");
|
||||
}
|
||||
}
|
||||
@@ -40,6 +40,7 @@ use tracing_subscriber::util::SubscriberInitExt;
|
||||
mod bespoke_event_handling;
|
||||
mod codex_message_processor;
|
||||
mod config_api;
|
||||
mod dynamic_tools;
|
||||
mod error_code;
|
||||
mod fuzzy_file_search;
|
||||
mod message_processor;
|
||||
|
||||
@@ -0,0 +1,286 @@
|
||||
use anyhow::Context;
|
||||
use anyhow::Result;
|
||||
use app_test_support::McpProcess;
|
||||
use app_test_support::create_final_assistant_message_sse_response;
|
||||
use app_test_support::create_mock_responses_server_sequence_unchecked;
|
||||
use app_test_support::to_response;
|
||||
use codex_app_server_protocol::DynamicToolCallParams;
|
||||
use codex_app_server_protocol::DynamicToolCallResponse;
|
||||
use codex_app_server_protocol::DynamicToolSpec;
|
||||
use codex_app_server_protocol::JSONRPCResponse;
|
||||
use codex_app_server_protocol::RequestId;
|
||||
use codex_app_server_protocol::ServerRequest;
|
||||
use codex_app_server_protocol::ThreadStartParams;
|
||||
use codex_app_server_protocol::ThreadStartResponse;
|
||||
use codex_app_server_protocol::TurnStartParams;
|
||||
use codex_app_server_protocol::TurnStartResponse;
|
||||
use codex_app_server_protocol::UserInput as V2UserInput;
|
||||
use core_test_support::responses;
|
||||
use pretty_assertions::assert_eq;
|
||||
use serde_json::Value;
|
||||
use serde_json::json;
|
||||
use std::path::Path;
|
||||
use std::time::Duration;
|
||||
use tempfile::TempDir;
|
||||
use tokio::time::timeout;
|
||||
use wiremock::MockServer;
|
||||
|
||||
const DEFAULT_READ_TIMEOUT: Duration = Duration::from_secs(10);
|
||||
|
||||
/// Ensures dynamic tool specs are serialized into the model request payload.
|
||||
#[tokio::test]
|
||||
async fn thread_start_injects_dynamic_tools_into_model_requests() -> Result<()> {
|
||||
let responses = vec![create_final_assistant_message_sse_response("Done")?];
|
||||
let server = create_mock_responses_server_sequence_unchecked(responses).await;
|
||||
|
||||
let codex_home = TempDir::new()?;
|
||||
create_config_toml(codex_home.path(), &server.uri())?;
|
||||
|
||||
let mut mcp = McpProcess::new(codex_home.path()).await?;
|
||||
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
|
||||
|
||||
// Use a minimal JSON schema so we can assert the tool payload round-trips.
|
||||
let input_schema = json!({
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"city": { "type": "string" }
|
||||
},
|
||||
"required": ["city"],
|
||||
"additionalProperties": false,
|
||||
});
|
||||
let dynamic_tool = DynamicToolSpec {
|
||||
name: "demo_tool".to_string(),
|
||||
description: "Demo dynamic tool".to_string(),
|
||||
input_schema: input_schema.clone(),
|
||||
};
|
||||
|
||||
// Thread start injects dynamic tools into the thread's tool registry.
|
||||
let thread_req = mcp
|
||||
.send_thread_start_request(ThreadStartParams {
|
||||
dynamic_tools: Some(vec![dynamic_tool.clone()]),
|
||||
..Default::default()
|
||||
})
|
||||
.await?;
|
||||
let thread_resp: JSONRPCResponse = timeout(
|
||||
DEFAULT_READ_TIMEOUT,
|
||||
mcp.read_stream_until_response_message(RequestId::Integer(thread_req)),
|
||||
)
|
||||
.await??;
|
||||
let ThreadStartResponse { thread, .. } = to_response::<ThreadStartResponse>(thread_resp)?;
|
||||
|
||||
// Start a turn so a model request is issued.
|
||||
let turn_req = mcp
|
||||
.send_turn_start_request(TurnStartParams {
|
||||
thread_id: thread.id.clone(),
|
||||
input: vec![V2UserInput::Text {
|
||||
text: "Hello".to_string(),
|
||||
text_elements: Vec::new(),
|
||||
}],
|
||||
..Default::default()
|
||||
})
|
||||
.await?;
|
||||
let turn_resp: JSONRPCResponse = timeout(
|
||||
DEFAULT_READ_TIMEOUT,
|
||||
mcp.read_stream_until_response_message(RequestId::Integer(turn_req)),
|
||||
)
|
||||
.await??;
|
||||
let _turn: TurnStartResponse = to_response::<TurnStartResponse>(turn_resp)?;
|
||||
|
||||
timeout(
|
||||
DEFAULT_READ_TIMEOUT,
|
||||
mcp.read_stream_until_notification_message("turn/completed"),
|
||||
)
|
||||
.await??;
|
||||
|
||||
// Inspect the captured model request to assert the tool spec made it through.
|
||||
let bodies = responses_bodies(&server).await?;
|
||||
let body = bodies
|
||||
.first()
|
||||
.context("expected at least one responses request")?;
|
||||
let tool = find_tool(body, &dynamic_tool.name)
|
||||
.context("expected dynamic tool to be injected into request")?;
|
||||
|
||||
assert_eq!(
|
||||
tool.get("description"),
|
||||
Some(&Value::String(dynamic_tool.description.clone()))
|
||||
);
|
||||
assert_eq!(tool.get("parameters"), Some(&input_schema));
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
/// Exercises the full dynamic tool call path (server request, client response, model output).
|
||||
#[tokio::test]
|
||||
async fn dynamic_tool_call_round_trip_sends_output_to_model() -> Result<()> {
|
||||
let call_id = "dyn-call-1";
|
||||
let tool_name = "demo_tool";
|
||||
let tool_args = json!({ "city": "Paris" });
|
||||
let tool_call_arguments = serde_json::to_string(&tool_args)?;
|
||||
|
||||
// First response triggers a dynamic tool call, second closes the turn.
|
||||
let responses = vec![
|
||||
responses::sse(vec![
|
||||
responses::ev_response_created("resp-1"),
|
||||
responses::ev_function_call(call_id, tool_name, &tool_call_arguments),
|
||||
responses::ev_completed("resp-1"),
|
||||
]),
|
||||
create_final_assistant_message_sse_response("Done")?,
|
||||
];
|
||||
let server = create_mock_responses_server_sequence_unchecked(responses).await;
|
||||
|
||||
let codex_home = TempDir::new()?;
|
||||
create_config_toml(codex_home.path(), &server.uri())?;
|
||||
|
||||
let mut mcp = McpProcess::new(codex_home.path()).await?;
|
||||
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
|
||||
|
||||
let dynamic_tool = DynamicToolSpec {
|
||||
name: tool_name.to_string(),
|
||||
description: "Demo dynamic tool".to_string(),
|
||||
input_schema: json!({
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"city": { "type": "string" }
|
||||
},
|
||||
"required": ["city"],
|
||||
"additionalProperties": false,
|
||||
}),
|
||||
};
|
||||
|
||||
let thread_req = mcp
|
||||
.send_thread_start_request(ThreadStartParams {
|
||||
dynamic_tools: Some(vec![dynamic_tool]),
|
||||
..Default::default()
|
||||
})
|
||||
.await?;
|
||||
let thread_resp: JSONRPCResponse = timeout(
|
||||
DEFAULT_READ_TIMEOUT,
|
||||
mcp.read_stream_until_response_message(RequestId::Integer(thread_req)),
|
||||
)
|
||||
.await??;
|
||||
let ThreadStartResponse { thread, .. } = to_response::<ThreadStartResponse>(thread_resp)?;
|
||||
|
||||
// Start a turn so the tool call is emitted.
|
||||
let turn_req = mcp
|
||||
.send_turn_start_request(TurnStartParams {
|
||||
thread_id: thread.id.clone(),
|
||||
input: vec![V2UserInput::Text {
|
||||
text: "Run the tool".to_string(),
|
||||
text_elements: Vec::new(),
|
||||
}],
|
||||
..Default::default()
|
||||
})
|
||||
.await?;
|
||||
let turn_resp: JSONRPCResponse = timeout(
|
||||
DEFAULT_READ_TIMEOUT,
|
||||
mcp.read_stream_until_response_message(RequestId::Integer(turn_req)),
|
||||
)
|
||||
.await??;
|
||||
let TurnStartResponse { turn } = to_response::<TurnStartResponse>(turn_resp)?;
|
||||
|
||||
// Read the tool call request from the app server.
|
||||
let request = timeout(
|
||||
DEFAULT_READ_TIMEOUT,
|
||||
mcp.read_stream_until_request_message(),
|
||||
)
|
||||
.await??;
|
||||
let (request_id, params) = match request {
|
||||
ServerRequest::DynamicToolCall { request_id, params } => (request_id, params),
|
||||
other => panic!("expected DynamicToolCall request, got {other:?}"),
|
||||
};
|
||||
|
||||
let expected = DynamicToolCallParams {
|
||||
thread_id: thread.id,
|
||||
turn_id: turn.id,
|
||||
call_id: call_id.to_string(),
|
||||
tool: tool_name.to_string(),
|
||||
arguments: tool_args.clone(),
|
||||
};
|
||||
assert_eq!(params, expected);
|
||||
|
||||
// Respond to the tool call so the model receives a function_call_output.
|
||||
let response = DynamicToolCallResponse {
|
||||
output: "dynamic-ok".to_string(),
|
||||
success: true,
|
||||
};
|
||||
mcp.send_response(request_id, serde_json::to_value(response)?)
|
||||
.await?;
|
||||
|
||||
timeout(
|
||||
DEFAULT_READ_TIMEOUT,
|
||||
mcp.read_stream_until_notification_message("turn/completed"),
|
||||
)
|
||||
.await??;
|
||||
|
||||
let bodies = responses_bodies(&server).await?;
|
||||
let output = bodies
|
||||
.iter()
|
||||
.find_map(|body| function_call_output_text(body, call_id))
|
||||
.context("expected function_call_output in follow-up request")?;
|
||||
assert_eq!(output, "dynamic-ok");
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
async fn responses_bodies(server: &MockServer) -> Result<Vec<Value>> {
|
||||
let requests = server
|
||||
.received_requests()
|
||||
.await
|
||||
.context("failed to fetch received requests")?;
|
||||
|
||||
requests
|
||||
.into_iter()
|
||||
.filter(|req| req.url.path().ends_with("/responses"))
|
||||
.map(|req| {
|
||||
req.body_json::<Value>()
|
||||
.context("request body should be JSON")
|
||||
})
|
||||
.collect()
|
||||
}
|
||||
|
||||
fn find_tool<'a>(body: &'a Value, name: &str) -> Option<&'a Value> {
|
||||
body.get("tools")
|
||||
.and_then(Value::as_array)
|
||||
.and_then(|tools| {
|
||||
tools
|
||||
.iter()
|
||||
.find(|tool| tool.get("name").and_then(Value::as_str) == Some(name))
|
||||
})
|
||||
}
|
||||
|
||||
fn function_call_output_text(body: &Value, call_id: &str) -> Option<String> {
|
||||
body.get("input")
|
||||
.and_then(Value::as_array)
|
||||
.and_then(|items| {
|
||||
items.iter().find(|item| {
|
||||
item.get("type").and_then(Value::as_str) == Some("function_call_output")
|
||||
&& item.get("call_id").and_then(Value::as_str) == Some(call_id)
|
||||
})
|
||||
})
|
||||
.and_then(|item| item.get("output"))
|
||||
.and_then(Value::as_str)
|
||||
.map(str::to_string)
|
||||
}
|
||||
|
||||
fn create_config_toml(codex_home: &Path, server_uri: &str) -> std::io::Result<()> {
|
||||
let config_toml = codex_home.join("config.toml");
|
||||
std::fs::write(
|
||||
config_toml,
|
||||
format!(
|
||||
r#"
|
||||
model = "mock-model"
|
||||
approval_policy = "never"
|
||||
sandbox_mode = "read-only"
|
||||
|
||||
model_provider = "mock_provider"
|
||||
|
||||
[model_providers.mock_provider]
|
||||
name = "Mock provider for test"
|
||||
base_url = "{server_uri}/v1"
|
||||
wire_api = "responses"
|
||||
request_max_retries = 0
|
||||
stream_max_retries = 0
|
||||
"#
|
||||
),
|
||||
)
|
||||
}
|
||||
@@ -3,6 +3,7 @@ mod analytics;
|
||||
mod app_list;
|
||||
mod collaboration_mode_list;
|
||||
mod config_rpc;
|
||||
mod dynamic_tools;
|
||||
mod initialize;
|
||||
mod model_list;
|
||||
mod output_schema;
|
||||
|
||||
@@ -38,6 +38,8 @@ use codex_protocol::approvals::ExecPolicyAmendment;
|
||||
use codex_protocol::config_types::ModeKind;
|
||||
use codex_protocol::config_types::Settings;
|
||||
use codex_protocol::config_types::WebSearchMode;
|
||||
use codex_protocol::dynamic_tools::DynamicToolResponse;
|
||||
use codex_protocol::dynamic_tools::DynamicToolSpec;
|
||||
use codex_protocol::items::TurnItem;
|
||||
use codex_protocol::items::UserMessageItem;
|
||||
use codex_protocol::models::BaseInstructions;
|
||||
@@ -236,6 +238,7 @@ fn maybe_push_chat_wire_api_deprecation(
|
||||
|
||||
impl Codex {
|
||||
/// Spawn a new [`Codex`] and initialize the session.
|
||||
#[allow(clippy::too_many_arguments)]
|
||||
pub(crate) async fn spawn(
|
||||
config: Config,
|
||||
auth_manager: Arc<AuthManager>,
|
||||
@@ -244,16 +247,12 @@ impl Codex {
|
||||
conversation_history: InitialHistory,
|
||||
session_source: SessionSource,
|
||||
agent_control: AgentControl,
|
||||
dynamic_tools: Vec<DynamicToolSpec>,
|
||||
) -> CodexResult<CodexSpawnOk> {
|
||||
let (tx_sub, rx_sub) = async_channel::bounded(SUBMISSION_CHANNEL_CAPACITY);
|
||||
let (tx_event, rx_event) = async_channel::unbounded();
|
||||
|
||||
let loaded_skills = skills_manager.skills_for_config(&config);
|
||||
// let loaded_skills = if config.features.enabled(Feature::Skills) {
|
||||
// Some(skills_manager.skills_for_config(&config))
|
||||
// } else {
|
||||
// None
|
||||
// };
|
||||
|
||||
for err in &loaded_skills.errors {
|
||||
error!(
|
||||
@@ -320,6 +319,7 @@ impl Codex {
|
||||
cwd: config.cwd.clone(),
|
||||
original_config_do_not_use: Arc::clone(&config),
|
||||
session_source,
|
||||
dynamic_tools,
|
||||
};
|
||||
|
||||
// Generate a unique ID for the lifetime of this Codex session.
|
||||
@@ -443,6 +443,7 @@ pub(crate) struct TurnContext {
|
||||
pub(crate) codex_linux_sandbox_exe: Option<PathBuf>,
|
||||
pub(crate) tool_call_gate: Arc<ReadinessFlag>,
|
||||
pub(crate) truncation_policy: TruncationPolicy,
|
||||
pub(crate) dynamic_tools: Vec<DynamicToolSpec>,
|
||||
}
|
||||
|
||||
impl TurnContext {
|
||||
@@ -500,6 +501,7 @@ pub(crate) struct SessionConfiguration {
|
||||
original_config_do_not_use: Arc<Config>,
|
||||
/// Source of the session (cli, vscode, exec, mcp, ...)
|
||||
session_source: SessionSource,
|
||||
dynamic_tools: Vec<DynamicToolSpec>,
|
||||
}
|
||||
|
||||
impl SessionConfiguration {
|
||||
@@ -616,6 +618,7 @@ impl Session {
|
||||
codex_linux_sandbox_exe: per_turn_config.codex_linux_sandbox_exe.clone(),
|
||||
tool_call_gate: Arc::new(ReadinessFlag::new()),
|
||||
truncation_policy: model_info.truncation_policy.into(),
|
||||
dynamic_tools: session_configuration.dynamic_tools.clone(),
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1495,6 +1498,27 @@ impl Session {
|
||||
}
|
||||
}
|
||||
|
||||
pub async fn notify_dynamic_tool_response(&self, call_id: &str, response: DynamicToolResponse) {
|
||||
let entry = {
|
||||
let mut active = self.active_turn.lock().await;
|
||||
match active.as_mut() {
|
||||
Some(at) => {
|
||||
let mut ts = at.turn_state.lock().await;
|
||||
ts.remove_pending_dynamic_tool(call_id)
|
||||
}
|
||||
None => None,
|
||||
}
|
||||
};
|
||||
match entry {
|
||||
Some(tx_response) => {
|
||||
tx_response.send(response).ok();
|
||||
}
|
||||
None => {
|
||||
warn!("No pending dynamic tool call found for call_id: {call_id}");
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
pub async fn notify_approval(&self, sub_id: &str, decision: ReviewDecision) {
|
||||
let entry = {
|
||||
let mut active = self.active_turn.lock().await;
|
||||
@@ -2156,6 +2180,9 @@ async fn submission_loop(sess: Arc<Session>, config: Arc<Config>, rx_sub: Receiv
|
||||
Op::UserInputAnswer { id, response } => {
|
||||
handlers::request_user_input_response(&sess, id, response).await;
|
||||
}
|
||||
Op::DynamicToolResponse { id, response } => {
|
||||
handlers::dynamic_tool_response(&sess, id, response).await;
|
||||
}
|
||||
Op::AddToHistory { text } => {
|
||||
handlers::add_to_history(&sess, &config, text).await;
|
||||
}
|
||||
@@ -2252,6 +2279,7 @@ mod handlers {
|
||||
use codex_protocol::config_types::CollaborationMode;
|
||||
use codex_protocol::config_types::ModeKind;
|
||||
use codex_protocol::config_types::Settings;
|
||||
use codex_protocol::dynamic_tools::DynamicToolResponse;
|
||||
use codex_protocol::user_input::UserInput;
|
||||
use codex_rmcp_client::ElicitationAction;
|
||||
use codex_rmcp_client::ElicitationResponse;
|
||||
@@ -2489,6 +2517,14 @@ mod handlers {
|
||||
sess.notify_user_input_response(&id, response).await;
|
||||
}
|
||||
|
||||
pub async fn dynamic_tool_response(
|
||||
sess: &Arc<Session>,
|
||||
id: String,
|
||||
response: DynamicToolResponse,
|
||||
) {
|
||||
sess.notify_dynamic_tool_response(&id, response).await;
|
||||
}
|
||||
|
||||
pub async fn add_to_history(sess: &Arc<Session>, config: &Arc<Config>, text: String) {
|
||||
let id = sess.conversation_id;
|
||||
let config = Arc::clone(config);
|
||||
@@ -2826,6 +2862,7 @@ async fn spawn_review_thread(
|
||||
final_output_json_schema: None,
|
||||
codex_linux_sandbox_exe: parent_turn_context.codex_linux_sandbox_exe.clone(),
|
||||
tool_call_gate: Arc::new(ReadinessFlag::new()),
|
||||
dynamic_tools: parent_turn_context.dynamic_tools.clone(),
|
||||
truncation_policy: model_info.truncation_policy.into(),
|
||||
};
|
||||
|
||||
@@ -3162,6 +3199,7 @@ async fn run_sampling_request(
|
||||
.map(|(name, tool)| (name, tool.tool))
|
||||
.collect(),
|
||||
),
|
||||
turn_context.dynamic_tools.as_slice(),
|
||||
));
|
||||
|
||||
let model_supports_parallel = turn_context
|
||||
@@ -3942,6 +3980,7 @@ mod tests {
|
||||
cwd: config.cwd.clone(),
|
||||
original_config_do_not_use: Arc::clone(&config),
|
||||
session_source: SessionSource::Exec,
|
||||
dynamic_tools: Vec::new(),
|
||||
};
|
||||
|
||||
let mut state = SessionState::new(session_configuration);
|
||||
@@ -4021,6 +4060,7 @@ mod tests {
|
||||
cwd: config.cwd.clone(),
|
||||
original_config_do_not_use: Arc::clone(&config),
|
||||
session_source: SessionSource::Exec,
|
||||
dynamic_tools: Vec::new(),
|
||||
};
|
||||
|
||||
let mut state = SessionState::new(session_configuration);
|
||||
@@ -4284,6 +4324,7 @@ mod tests {
|
||||
cwd: config.cwd.clone(),
|
||||
original_config_do_not_use: Arc::clone(&config),
|
||||
session_source: SessionSource::Exec,
|
||||
dynamic_tools: Vec::new(),
|
||||
};
|
||||
let per_turn_config = Session::build_per_turn_config(&session_configuration);
|
||||
let model_info = ModelsManager::construct_model_info_offline(
|
||||
@@ -4392,6 +4433,7 @@ mod tests {
|
||||
cwd: config.cwd.clone(),
|
||||
original_config_do_not_use: Arc::clone(&config),
|
||||
session_source: SessionSource::Exec,
|
||||
dynamic_tools: Vec::new(),
|
||||
};
|
||||
let per_turn_config = Session::build_per_turn_config(&session_configuration);
|
||||
let model_info = ModelsManager::construct_model_info_offline(
|
||||
@@ -4712,6 +4754,7 @@ mod tests {
|
||||
.map(|(name, tool)| (name, tool.tool))
|
||||
.collect(),
|
||||
),
|
||||
turn_context.dynamic_tools.as_slice(),
|
||||
);
|
||||
let item = ResponseItem::CustomToolCall {
|
||||
id: None,
|
||||
|
||||
@@ -57,6 +57,7 @@ pub(crate) async fn run_codex_thread_interactive(
|
||||
initial_history.unwrap_or(InitialHistory::New),
|
||||
SessionSource::SubAgent(SubAgentSource::Review),
|
||||
parent_session.services.agent_control.clone(),
|
||||
Vec::new(),
|
||||
)
|
||||
.await?;
|
||||
let codex = Arc::new(codex);
|
||||
|
||||
@@ -126,6 +126,7 @@ pub use safety::get_platform_sandbox;
|
||||
pub use safety::is_windows_elevated_sandbox_enabled;
|
||||
pub use safety::set_windows_elevated_sandbox_enabled;
|
||||
pub use safety::set_windows_sandbox_enabled;
|
||||
pub use tools::spec::parse_tool_input_schema;
|
||||
// Re-export the protocol types from the standalone `codex-protocol` crate so existing
|
||||
// `codex_core::protocol::...` references continue to work across the workspace.
|
||||
pub use codex_protocol::protocol;
|
||||
|
||||
@@ -68,6 +68,7 @@ pub(crate) fn should_persist_event_msg(ev: &EventMsg) -> bool {
|
||||
| EventMsg::ExecCommandEnd(_)
|
||||
| EventMsg::ExecApprovalRequest(_)
|
||||
| EventMsg::RequestUserInput(_)
|
||||
| EventMsg::DynamicToolCallRequest(_)
|
||||
| EventMsg::ElicitationRequest(_)
|
||||
| EventMsg::ApplyPatchApprovalRequest(_)
|
||||
| EventMsg::BackgroundEvent(_)
|
||||
|
||||
@@ -8,6 +8,7 @@ use tokio::sync::Notify;
|
||||
use tokio_util::sync::CancellationToken;
|
||||
use tokio_util::task::AbortOnDropHandle;
|
||||
|
||||
use codex_protocol::dynamic_tools::DynamicToolResponse;
|
||||
use codex_protocol::models::ResponseInputItem;
|
||||
use codex_protocol::request_user_input::RequestUserInputResponse;
|
||||
use tokio::sync::oneshot;
|
||||
@@ -70,6 +71,7 @@ impl ActiveTurn {
|
||||
pub(crate) struct TurnState {
|
||||
pending_approvals: HashMap<String, oneshot::Sender<ReviewDecision>>,
|
||||
pending_user_input: HashMap<String, oneshot::Sender<RequestUserInputResponse>>,
|
||||
pending_dynamic_tools: HashMap<String, oneshot::Sender<DynamicToolResponse>>,
|
||||
pending_input: Vec<ResponseInputItem>,
|
||||
}
|
||||
|
||||
@@ -92,6 +94,7 @@ impl TurnState {
|
||||
pub(crate) fn clear_pending(&mut self) {
|
||||
self.pending_approvals.clear();
|
||||
self.pending_user_input.clear();
|
||||
self.pending_dynamic_tools.clear();
|
||||
self.pending_input.clear();
|
||||
}
|
||||
|
||||
@@ -110,6 +113,21 @@ impl TurnState {
|
||||
self.pending_user_input.remove(key)
|
||||
}
|
||||
|
||||
pub(crate) fn insert_pending_dynamic_tool(
|
||||
&mut self,
|
||||
key: String,
|
||||
tx: oneshot::Sender<DynamicToolResponse>,
|
||||
) -> Option<oneshot::Sender<DynamicToolResponse>> {
|
||||
self.pending_dynamic_tools.insert(key, tx)
|
||||
}
|
||||
|
||||
pub(crate) fn remove_pending_dynamic_tool(
|
||||
&mut self,
|
||||
key: &str,
|
||||
) -> Option<oneshot::Sender<DynamicToolResponse>> {
|
||||
self.pending_dynamic_tools.remove(key)
|
||||
}
|
||||
|
||||
pub(crate) fn push_pending_input(&mut self, input: ResponseInputItem) {
|
||||
self.pending_input.push(input);
|
||||
}
|
||||
|
||||
@@ -196,12 +196,21 @@ impl ThreadManager {
|
||||
}
|
||||
|
||||
pub async fn start_thread(&self, config: Config) -> CodexResult<NewThread> {
|
||||
self.start_thread_with_tools(config, Vec::new()).await
|
||||
}
|
||||
|
||||
pub async fn start_thread_with_tools(
|
||||
&self,
|
||||
config: Config,
|
||||
dynamic_tools: Vec<codex_protocol::dynamic_tools::DynamicToolSpec>,
|
||||
) -> CodexResult<NewThread> {
|
||||
self.state
|
||||
.spawn_thread(
|
||||
config,
|
||||
InitialHistory::New,
|
||||
Arc::clone(&self.state.auth_manager),
|
||||
self.agent_control(),
|
||||
dynamic_tools,
|
||||
)
|
||||
.await
|
||||
}
|
||||
@@ -224,7 +233,13 @@ impl ThreadManager {
|
||||
auth_manager: Arc<AuthManager>,
|
||||
) -> CodexResult<NewThread> {
|
||||
self.state
|
||||
.spawn_thread(config, initial_history, auth_manager, self.agent_control())
|
||||
.spawn_thread(
|
||||
config,
|
||||
initial_history,
|
||||
auth_manager,
|
||||
self.agent_control(),
|
||||
Vec::new(),
|
||||
)
|
||||
.await
|
||||
}
|
||||
|
||||
@@ -262,6 +277,7 @@ impl ThreadManager {
|
||||
history,
|
||||
Arc::clone(&self.state.auth_manager),
|
||||
self.agent_control(),
|
||||
Vec::new(),
|
||||
)
|
||||
.await
|
||||
}
|
||||
@@ -330,6 +346,7 @@ impl ThreadManagerState {
|
||||
Arc::clone(&self.auth_manager),
|
||||
agent_control,
|
||||
session_source,
|
||||
Vec::new(),
|
||||
)
|
||||
.await
|
||||
}
|
||||
@@ -341,6 +358,7 @@ impl ThreadManagerState {
|
||||
initial_history: InitialHistory,
|
||||
auth_manager: Arc<AuthManager>,
|
||||
agent_control: AgentControl,
|
||||
dynamic_tools: Vec<codex_protocol::dynamic_tools::DynamicToolSpec>,
|
||||
) -> CodexResult<NewThread> {
|
||||
self.spawn_thread_with_source(
|
||||
config,
|
||||
@@ -348,6 +366,7 @@ impl ThreadManagerState {
|
||||
auth_manager,
|
||||
agent_control,
|
||||
self.session_source.clone(),
|
||||
dynamic_tools,
|
||||
)
|
||||
.await
|
||||
}
|
||||
@@ -359,6 +378,7 @@ impl ThreadManagerState {
|
||||
auth_manager: Arc<AuthManager>,
|
||||
agent_control: AgentControl,
|
||||
session_source: SessionSource,
|
||||
dynamic_tools: Vec<codex_protocol::dynamic_tools::DynamicToolSpec>,
|
||||
) -> CodexResult<NewThread> {
|
||||
let CodexSpawnOk {
|
||||
codex, thread_id, ..
|
||||
@@ -370,6 +390,7 @@ impl ThreadManagerState {
|
||||
initial_history,
|
||||
session_source,
|
||||
agent_control,
|
||||
dynamic_tools,
|
||||
)
|
||||
.await?;
|
||||
self.finalize_thread_spawn(codex, thread_id).await
|
||||
|
||||
@@ -0,0 +1,98 @@
|
||||
use crate::codex::Session;
|
||||
use crate::codex::TurnContext;
|
||||
use crate::function_tool::FunctionCallError;
|
||||
use crate::tools::context::ToolInvocation;
|
||||
use crate::tools::context::ToolOutput;
|
||||
use crate::tools::context::ToolPayload;
|
||||
use crate::tools::handlers::parse_arguments;
|
||||
use crate::tools::registry::ToolHandler;
|
||||
use crate::tools::registry::ToolKind;
|
||||
use async_trait::async_trait;
|
||||
use codex_protocol::dynamic_tools::DynamicToolCallRequest;
|
||||
use codex_protocol::dynamic_tools::DynamicToolResponse;
|
||||
use codex_protocol::protocol::EventMsg;
|
||||
use serde_json::Value;
|
||||
use tokio::sync::oneshot;
|
||||
use tracing::warn;
|
||||
|
||||
pub struct DynamicToolHandler;
|
||||
|
||||
#[async_trait]
|
||||
impl ToolHandler for DynamicToolHandler {
|
||||
fn kind(&self) -> ToolKind {
|
||||
ToolKind::Function
|
||||
}
|
||||
|
||||
async fn is_mutating(&self, _invocation: &ToolInvocation) -> bool {
|
||||
true
|
||||
}
|
||||
|
||||
async fn handle(&self, invocation: ToolInvocation) -> Result<ToolOutput, FunctionCallError> {
|
||||
let ToolInvocation {
|
||||
session,
|
||||
turn,
|
||||
call_id,
|
||||
tool_name,
|
||||
payload,
|
||||
..
|
||||
} = invocation;
|
||||
|
||||
let arguments = match payload {
|
||||
ToolPayload::Function { arguments } => arguments,
|
||||
_ => {
|
||||
return Err(FunctionCallError::RespondToModel(
|
||||
"dynamic tool handler received unsupported payload".to_string(),
|
||||
));
|
||||
}
|
||||
};
|
||||
|
||||
let args: Value = parse_arguments(&arguments)?;
|
||||
let response = request_dynamic_tool(&session, turn.as_ref(), call_id, tool_name, args)
|
||||
.await
|
||||
.ok_or_else(|| {
|
||||
FunctionCallError::RespondToModel(
|
||||
"dynamic tool call was cancelled before receiving a response".to_string(),
|
||||
)
|
||||
})?;
|
||||
|
||||
Ok(ToolOutput::Function {
|
||||
content: response.output,
|
||||
content_items: None,
|
||||
success: Some(response.success),
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
async fn request_dynamic_tool(
|
||||
session: &Session,
|
||||
turn_context: &TurnContext,
|
||||
call_id: String,
|
||||
tool: String,
|
||||
arguments: Value,
|
||||
) -> Option<DynamicToolResponse> {
|
||||
let _sub_id = turn_context.sub_id.clone();
|
||||
let (tx_response, rx_response) = oneshot::channel();
|
||||
let event_id = call_id.clone();
|
||||
let prev_entry = {
|
||||
let mut active = session.active_turn.lock().await;
|
||||
match active.as_mut() {
|
||||
Some(at) => {
|
||||
let mut ts = at.turn_state.lock().await;
|
||||
ts.insert_pending_dynamic_tool(call_id.clone(), tx_response)
|
||||
}
|
||||
None => None,
|
||||
}
|
||||
};
|
||||
if prev_entry.is_some() {
|
||||
warn!("Overwriting existing pending dynamic tool call for call_id: {event_id}");
|
||||
}
|
||||
|
||||
let event = EventMsg::DynamicToolCallRequest(DynamicToolCallRequest {
|
||||
call_id,
|
||||
turn_id: turn_context.sub_id.clone(),
|
||||
tool,
|
||||
arguments,
|
||||
});
|
||||
session.send_event(turn_context, event).await;
|
||||
rx_response.await.ok()
|
||||
}
|
||||
@@ -1,5 +1,6 @@
|
||||
pub mod apply_patch;
|
||||
pub(crate) mod collab;
|
||||
mod dynamic;
|
||||
mod grep_files;
|
||||
mod list_dir;
|
||||
mod mcp;
|
||||
@@ -18,6 +19,7 @@ use serde::Deserialize;
|
||||
use crate::function_tool::FunctionCallError;
|
||||
pub use apply_patch::ApplyPatchHandler;
|
||||
pub use collab::CollabHandler;
|
||||
pub use dynamic::DynamicToolHandler;
|
||||
pub use grep_files::GrepFilesHandler;
|
||||
pub use list_dir::ListDirHandler;
|
||||
pub use mcp::McpHandler;
|
||||
|
||||
@@ -10,6 +10,7 @@ use crate::tools::registry::ConfiguredToolSpec;
|
||||
use crate::tools::registry::ToolRegistry;
|
||||
use crate::tools::spec::ToolsConfig;
|
||||
use crate::tools::spec::build_specs;
|
||||
use codex_protocol::dynamic_tools::DynamicToolSpec;
|
||||
use codex_protocol::models::LocalShellAction;
|
||||
use codex_protocol::models::ResponseInputItem;
|
||||
use codex_protocol::models::ResponseItem;
|
||||
@@ -34,8 +35,9 @@ impl ToolRouter {
|
||||
pub fn from_config(
|
||||
config: &ToolsConfig,
|
||||
mcp_tools: Option<HashMap<String, mcp_types::Tool>>,
|
||||
dynamic_tools: &[DynamicToolSpec],
|
||||
) -> Self {
|
||||
let builder = build_specs(config, mcp_tools);
|
||||
let builder = build_specs(config, mcp_tools, dynamic_tools);
|
||||
let (specs, registry) = builder.build();
|
||||
|
||||
Self { registry, specs }
|
||||
|
||||
@@ -10,6 +10,7 @@ use crate::tools::handlers::collab::DEFAULT_WAIT_TIMEOUT_MS;
|
||||
use crate::tools::handlers::collab::MAX_WAIT_TIMEOUT_MS;
|
||||
use crate::tools::registry::ToolRegistryBuilder;
|
||||
use codex_protocol::config_types::WebSearchMode;
|
||||
use codex_protocol::dynamic_tools::DynamicToolSpec;
|
||||
use codex_protocol::models::VIEW_IMAGE_TOOL_NAME;
|
||||
use codex_protocol::openai_models::ApplyPatchToolType;
|
||||
use codex_protocol::openai_models::ConfigShellToolType;
|
||||
@@ -87,7 +88,7 @@ impl ToolsConfig {
|
||||
/// Generic JSON‑Schema subset needed for our tool definitions
|
||||
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
|
||||
#[serde(tag = "type", rename_all = "lowercase")]
|
||||
pub(crate) enum JsonSchema {
|
||||
pub enum JsonSchema {
|
||||
Boolean {
|
||||
#[serde(skip_serializing_if = "Option::is_none")]
|
||||
description: Option<String>,
|
||||
@@ -123,7 +124,7 @@ pub(crate) enum JsonSchema {
|
||||
/// Whether additional properties are allowed, and if so, any required schema
|
||||
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
|
||||
#[serde(untagged)]
|
||||
pub(crate) enum AdditionalProperties {
|
||||
pub enum AdditionalProperties {
|
||||
Boolean(bool),
|
||||
Schema(Box<JsonSchema>),
|
||||
}
|
||||
@@ -1101,6 +1102,26 @@ pub(crate) fn mcp_tool_to_openai_tool(
|
||||
})
|
||||
}
|
||||
|
||||
fn dynamic_tool_to_openai_tool(
|
||||
tool: &DynamicToolSpec,
|
||||
) -> Result<ResponsesApiTool, serde_json::Error> {
|
||||
let input_schema = parse_tool_input_schema(&tool.input_schema)?;
|
||||
|
||||
Ok(ResponsesApiTool {
|
||||
name: tool.name.clone(),
|
||||
description: tool.description.clone(),
|
||||
strict: false,
|
||||
parameters: input_schema,
|
||||
})
|
||||
}
|
||||
|
||||
/// Parse the tool input_schema or return an error for invalid schema
|
||||
pub fn parse_tool_input_schema(input_schema: &JsonValue) -> Result<JsonSchema, serde_json::Error> {
|
||||
let mut input_schema = input_schema.clone();
|
||||
sanitize_json_schema(&mut input_schema);
|
||||
serde_json::from_value::<JsonSchema>(input_schema)
|
||||
}
|
||||
|
||||
/// Sanitize a JSON Schema (as serde_json::Value) so it can fit our limited
|
||||
/// JsonSchema enum. This function:
|
||||
/// - Ensures every schema object has a "type". If missing, infers it from
|
||||
@@ -1216,9 +1237,11 @@ fn sanitize_json_schema(value: &mut JsonValue) {
|
||||
pub(crate) fn build_specs(
|
||||
config: &ToolsConfig,
|
||||
mcp_tools: Option<HashMap<String, mcp_types::Tool>>,
|
||||
dynamic_tools: &[DynamicToolSpec],
|
||||
) -> ToolRegistryBuilder {
|
||||
use crate::tools::handlers::ApplyPatchHandler;
|
||||
use crate::tools::handlers::CollabHandler;
|
||||
use crate::tools::handlers::DynamicToolHandler;
|
||||
use crate::tools::handlers::GrepFilesHandler;
|
||||
use crate::tools::handlers::ListDirHandler;
|
||||
use crate::tools::handlers::McpHandler;
|
||||
@@ -1239,6 +1262,7 @@ pub(crate) fn build_specs(
|
||||
let unified_exec_handler = Arc::new(UnifiedExecHandler);
|
||||
let plan_handler = Arc::new(PlanHandler);
|
||||
let apply_patch_handler = Arc::new(ApplyPatchHandler);
|
||||
let dynamic_tool_handler = Arc::new(DynamicToolHandler);
|
||||
let view_image_handler = Arc::new(ViewImageHandler);
|
||||
let mcp_handler = Arc::new(McpHandler);
|
||||
let mcp_resource_handler = Arc::new(McpResourceHandler);
|
||||
@@ -1384,6 +1408,23 @@ pub(crate) fn build_specs(
|
||||
}
|
||||
}
|
||||
|
||||
if !dynamic_tools.is_empty() {
|
||||
for tool in dynamic_tools {
|
||||
match dynamic_tool_to_openai_tool(tool) {
|
||||
Ok(converted_tool) => {
|
||||
builder.push_spec(ToolSpec::Function(converted_tool));
|
||||
builder.register_handler(tool.name.clone(), dynamic_tool_handler.clone());
|
||||
}
|
||||
Err(e) => {
|
||||
tracing::error!(
|
||||
"Failed to convert dynamic tool {:?} to OpenAI tool: {e:?}",
|
||||
tool.name
|
||||
);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
builder
|
||||
}
|
||||
|
||||
@@ -1496,7 +1537,7 @@ mod tests {
|
||||
features: &features,
|
||||
web_search_mode: Some(WebSearchMode::Live),
|
||||
});
|
||||
let (tools, _) = build_specs(&config, None).build();
|
||||
let (tools, _) = build_specs(&config, None, &[]).build();
|
||||
|
||||
// Build actual map name -> spec
|
||||
use std::collections::BTreeMap;
|
||||
@@ -1560,7 +1601,7 @@ mod tests {
|
||||
features: &features,
|
||||
web_search_mode: Some(WebSearchMode::Cached),
|
||||
});
|
||||
let (tools, _) = build_specs(&tools_config, None).build();
|
||||
let (tools, _) = build_specs(&tools_config, None, &[]).build();
|
||||
assert_contains_tool_names(
|
||||
&tools,
|
||||
&["spawn_agent", "send_input", "wait", "close_agent"],
|
||||
@@ -1578,7 +1619,7 @@ mod tests {
|
||||
features: &features,
|
||||
web_search_mode: Some(WebSearchMode::Cached),
|
||||
});
|
||||
let (tools, _) = build_specs(&tools_config, None).build();
|
||||
let (tools, _) = build_specs(&tools_config, None, &[]).build();
|
||||
assert!(
|
||||
!tools.iter().any(|t| t.spec.name() == "request_user_input"),
|
||||
"request_user_input should be disabled when collaboration_modes feature is off"
|
||||
@@ -1590,7 +1631,7 @@ mod tests {
|
||||
features: &features,
|
||||
web_search_mode: Some(WebSearchMode::Cached),
|
||||
});
|
||||
let (tools, _) = build_specs(&tools_config, None).build();
|
||||
let (tools, _) = build_specs(&tools_config, None, &[]).build();
|
||||
assert_contains_tool_names(&tools, &["request_user_input"]);
|
||||
}
|
||||
|
||||
@@ -1607,7 +1648,7 @@ mod tests {
|
||||
features,
|
||||
web_search_mode,
|
||||
});
|
||||
let (tools, _) = build_specs(&tools_config, Some(HashMap::new())).build();
|
||||
let (tools, _) = build_specs(&tools_config, Some(HashMap::new()), &[]).build();
|
||||
let tool_names = tools.iter().map(|t| t.spec.name()).collect::<Vec<_>>();
|
||||
assert_eq!(&tool_names, &expected_tools,);
|
||||
}
|
||||
@@ -1623,7 +1664,7 @@ mod tests {
|
||||
features: &features,
|
||||
web_search_mode: Some(WebSearchMode::Cached),
|
||||
});
|
||||
let (tools, _) = build_specs(&tools_config, None).build();
|
||||
let (tools, _) = build_specs(&tools_config, None, &[]).build();
|
||||
|
||||
let tool = find_tool(&tools, "web_search");
|
||||
assert_eq!(
|
||||
@@ -1645,7 +1686,7 @@ mod tests {
|
||||
features: &features,
|
||||
web_search_mode: Some(WebSearchMode::Live),
|
||||
});
|
||||
let (tools, _) = build_specs(&tools_config, None).build();
|
||||
let (tools, _) = build_specs(&tools_config, None, &[]).build();
|
||||
|
||||
let tool = find_tool(&tools, "web_search");
|
||||
assert_eq!(
|
||||
@@ -1891,7 +1932,7 @@ mod tests {
|
||||
features: &features,
|
||||
web_search_mode: Some(WebSearchMode::Live),
|
||||
});
|
||||
let (tools, _) = build_specs(&tools_config, Some(HashMap::new())).build();
|
||||
let (tools, _) = build_specs(&tools_config, Some(HashMap::new()), &[]).build();
|
||||
|
||||
// Only check the shell variant and a couple of core tools.
|
||||
let mut subset = vec!["exec_command", "write_stdin", "update_plan"];
|
||||
@@ -1913,7 +1954,7 @@ mod tests {
|
||||
features: &features,
|
||||
web_search_mode: Some(WebSearchMode::Cached),
|
||||
});
|
||||
let (tools, _) = build_specs(&tools_config, None).build();
|
||||
let (tools, _) = build_specs(&tools_config, None, &[]).build();
|
||||
|
||||
assert!(!find_tool(&tools, "exec_command").supports_parallel_tool_calls);
|
||||
assert!(!find_tool(&tools, "write_stdin").supports_parallel_tool_calls);
|
||||
@@ -1932,7 +1973,7 @@ mod tests {
|
||||
features: &features,
|
||||
web_search_mode: Some(WebSearchMode::Cached),
|
||||
});
|
||||
let (tools, _) = build_specs(&tools_config, None).build();
|
||||
let (tools, _) = build_specs(&tools_config, None, &[]).build();
|
||||
|
||||
assert!(
|
||||
tools
|
||||
@@ -1999,6 +2040,7 @@ mod tests {
|
||||
description: Some("Do something cool".to_string()),
|
||||
},
|
||||
)])),
|
||||
&[],
|
||||
)
|
||||
.build();
|
||||
|
||||
@@ -2108,7 +2150,7 @@ mod tests {
|
||||
),
|
||||
]);
|
||||
|
||||
let (tools, _) = build_specs(&tools_config, Some(tools_map)).build();
|
||||
let (tools, _) = build_specs(&tools_config, Some(tools_map), &[]).build();
|
||||
|
||||
// Only assert that the MCP tools themselves are sorted by fully-qualified name.
|
||||
let mcp_names: Vec<_> = tools
|
||||
@@ -2157,6 +2199,7 @@ mod tests {
|
||||
description: Some("Search docs".to_string()),
|
||||
},
|
||||
)])),
|
||||
&[],
|
||||
)
|
||||
.build();
|
||||
|
||||
@@ -2212,6 +2255,7 @@ mod tests {
|
||||
description: Some("Pagination".to_string()),
|
||||
},
|
||||
)])),
|
||||
&[],
|
||||
)
|
||||
.build();
|
||||
|
||||
@@ -2266,6 +2310,7 @@ mod tests {
|
||||
description: Some("Tags".to_string()),
|
||||
},
|
||||
)])),
|
||||
&[],
|
||||
)
|
||||
.build();
|
||||
|
||||
@@ -2322,6 +2367,7 @@ mod tests {
|
||||
description: Some("AnyOf Value".to_string()),
|
||||
},
|
||||
)])),
|
||||
&[],
|
||||
)
|
||||
.build();
|
||||
|
||||
@@ -2459,6 +2505,7 @@ Examples of valid command strings:
|
||||
description: Some("Do something cool".to_string()),
|
||||
},
|
||||
)])),
|
||||
&[],
|
||||
)
|
||||
.build();
|
||||
|
||||
|
||||
@@ -607,7 +607,8 @@ impl EventProcessor for EventProcessorWithHumanOutput {
|
||||
| EventMsg::UndoCompleted(_)
|
||||
| EventMsg::UndoStarted(_)
|
||||
| EventMsg::ThreadRolledBack(_)
|
||||
| EventMsg::RequestUserInput(_) => {}
|
||||
| EventMsg::RequestUserInput(_)
|
||||
| EventMsg::DynamicToolCallRequest(_) => {}
|
||||
}
|
||||
CodexStatus::Running
|
||||
}
|
||||
|
||||
@@ -360,6 +360,7 @@ async fn run_codex_tool_session_inner(
|
||||
| EventMsg::UndoCompleted(_)
|
||||
| EventMsg::ExitedReviewMode(_)
|
||||
| EventMsg::RequestUserInput(_)
|
||||
| EventMsg::DynamicToolCallRequest(_)
|
||||
| EventMsg::ContextCompacted(_)
|
||||
| EventMsg::ThreadRolledBack(_)
|
||||
| EventMsg::CollabAgentSpawnBegin(_)
|
||||
|
||||
@@ -0,0 +1,30 @@
|
||||
use schemars::JsonSchema;
|
||||
use serde::Deserialize;
|
||||
use serde::Serialize;
|
||||
use serde_json::Value as JsonValue;
|
||||
use ts_rs::TS;
|
||||
|
||||
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, JsonSchema, TS)]
|
||||
#[serde(rename_all = "camelCase")]
|
||||
pub struct DynamicToolSpec {
|
||||
pub name: String,
|
||||
pub description: String,
|
||||
pub input_schema: JsonValue,
|
||||
}
|
||||
|
||||
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, JsonSchema, TS)]
|
||||
#[serde(rename_all = "camelCase")]
|
||||
pub struct DynamicToolCallRequest {
|
||||
pub call_id: String,
|
||||
pub turn_id: String,
|
||||
pub tool: String,
|
||||
pub arguments: JsonValue,
|
||||
}
|
||||
|
||||
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, JsonSchema, TS)]
|
||||
#[serde(rename_all = "camelCase")]
|
||||
pub struct DynamicToolResponse {
|
||||
pub call_id: String,
|
||||
pub output: String,
|
||||
pub success: bool,
|
||||
}
|
||||
@@ -4,6 +4,7 @@ pub use thread_id::ThreadId;
|
||||
pub mod approvals;
|
||||
pub mod config_types;
|
||||
pub mod custom_prompts;
|
||||
pub mod dynamic_tools;
|
||||
pub mod items;
|
||||
pub mod message_history;
|
||||
pub mod models;
|
||||
|
||||
@@ -17,6 +17,8 @@ use crate::config_types::CollaborationMode;
|
||||
use crate::config_types::Personality;
|
||||
use crate::config_types::ReasoningSummary as ReasoningSummaryConfig;
|
||||
use crate::custom_prompts::CustomPrompt;
|
||||
use crate::dynamic_tools::DynamicToolCallRequest;
|
||||
use crate::dynamic_tools::DynamicToolResponse;
|
||||
use crate::items::TurnItem;
|
||||
use crate::message_history::HistoryEntry;
|
||||
use crate::models::BaseInstructions;
|
||||
@@ -216,6 +218,14 @@ pub enum Op {
|
||||
response: RequestUserInputResponse,
|
||||
},
|
||||
|
||||
/// Resolve a dynamic tool call request.
|
||||
DynamicToolResponse {
|
||||
/// Call id for the in-flight request.
|
||||
id: String,
|
||||
/// Tool output payload.
|
||||
response: DynamicToolResponse,
|
||||
},
|
||||
|
||||
/// Append an entry to the persistent cross-session message history.
|
||||
///
|
||||
/// Note the entry is not guaranteed to be logged if the user has
|
||||
@@ -750,6 +760,8 @@ pub enum EventMsg {
|
||||
|
||||
RequestUserInput(RequestUserInputEvent),
|
||||
|
||||
DynamicToolCallRequest(DynamicToolCallRequest),
|
||||
|
||||
ElicitationRequest(ElicitationRequestEvent),
|
||||
|
||||
ApplyPatchApprovalRequest(ApplyPatchApprovalRequestEvent),
|
||||
|
||||
@@ -3043,7 +3043,8 @@ impl ChatWidget {
|
||||
| EventMsg::ItemCompleted(_)
|
||||
| EventMsg::AgentMessageContentDelta(_)
|
||||
| EventMsg::ReasoningContentDelta(_)
|
||||
| EventMsg::ReasoningRawContentDelta(_) => {}
|
||||
| EventMsg::ReasoningRawContentDelta(_)
|
||||
| EventMsg::DynamicToolCallRequest(_) => {}
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user