Implement recording functionality in agent test commands

- Added RecordingState structure to manage recording session details.
- Implemented SaveRecordingState and LoadRecordingState functions to handle recording state persistence.
- Enhanced HandleTestRecordCommand to support starting and stopping recording sessions with various options.
- Integrated gRPC calls for starting and stopping recordings in GuiAutomationClient.
- Updated ProposalRegistry to include sandbox directory and ROM path in ProposalMetadata.
- Implemented JSON parsing for Tile16Proposal to handle proposal creation from JSON input.
- Updated CMakeLists.txt to ensure proper linking of emulator targets.
This commit is contained in:
scawful
2025-10-03 21:16:08 -04:00
parent 5419633c52
commit 2c45453dd0
9 changed files with 973 additions and 109 deletions

View File

@@ -75,7 +75,8 @@
"YAZE_TEST_ROM_PATH": "${sourceDir}/zelda3.sfc",
"YAZE_WITH_JSON": "ON",
"YAZE_WITH_GRPC": "ON",
"YAZE_BUILD_Z3ED": "ON"
"YAZE_BUILD_Z3ED": "ON",
"YAZE_BUILD_EMU": "ON"
}
},
{

View File

@@ -1,8 +1,10 @@
#include "cli/handlers/agent/commands.h"
#include <algorithm>
#include <filesystem>
#include <fstream>
#include <optional>
#include <sstream>
#include <string>
#include <utility>
#include <vector>
@@ -154,7 +156,8 @@ absl::Status HandleRunCommand(const std::vector<std::string>& arg_vec,
if (!response_or.ok()) {
return response_or.status();
}
std::vector<std::string> commands = response_or.value().commands;
AgentResponse response = std::move(response_or.value());
const std::vector<std::string>& commands = response.commands;
// 3. Generate a structured proposal from the commands
Tile16ProposalGenerator generator;
@@ -186,11 +189,77 @@ absl::Status HandleRunCommand(const std::vector<std::string>& arg_vec,
absl::StrCat("Failed to save sandbox ROM: ", save_status.message()));
}
// 6. Save the proposal metadata for later use (accept/reject)
// For now, we'll just use the proposal generator's save function.
// A better approach would be to integrate with ProposalRegistry.
auto proposal_path =
RomSandboxManager::Instance().RootDirectory() / (proposal.id + ".json");
// 6. Persist the proposal metadata and artifacts.
auto& registry = ProposalRegistry::Instance();
int executed_commands = 0;
for (const auto& command : commands) {
if (command.empty() || command[0] == '#') {
continue;
}
++executed_commands;
}
std::string description = absl::StrFormat(
"Tile16 overworld edits (%d change%s)", proposal.changes.size(),
proposal.changes.size() == 1 ? "" : "s");
ASSIGN_OR_RETURN(
auto metadata,
registry.CreateProposal(sandbox.id, sandbox.rom_path, prompt, description));
proposal.id = metadata.id;
std::ostringstream diff_stream;
diff_stream << "Tile16 Proposal ID: " << metadata.id << "\n";
diff_stream << "Sandbox ID: " << sandbox.id << "\n";
diff_stream << "Sandbox ROM: " << sandbox.rom_path << "\n\n";
diff_stream << "Changes (" << proposal.changes.size() << "):\n";
for (const auto& change : proposal.changes) {
diff_stream << " - " << change.ToString() << "\n";
}
RETURN_IF_ERROR(registry.RecordDiff(metadata.id, diff_stream.str()));
RETURN_IF_ERROR(registry.AppendLog(
metadata.id, absl::StrCat("Prompt: ", prompt)));
if (!response.text_response.empty()) {
RETURN_IF_ERROR(registry.AppendLog(
metadata.id, absl::StrCat("AI Response: ", response.text_response)));
}
if (!response.reasoning.empty()) {
RETURN_IF_ERROR(registry.AppendLog(
metadata.id, absl::StrCat("Reasoning: ", response.reasoning)));
}
if (!response.tool_calls.empty()) {
RETURN_IF_ERROR(registry.AppendLog(
metadata.id,
absl::StrCat("Tool Calls: ", response.tool_calls.size())));
}
for (const auto& command : commands) {
if (command.empty() || command[0] == '#') {
continue;
}
RETURN_IF_ERROR(registry.AppendLog(
metadata.id, absl::StrCat("Command: ", command)));
}
RETURN_IF_ERROR(registry.AppendLog(
metadata.id,
absl::StrCat("Sandbox ROM saved to ", sandbox.rom_path.string())));
RETURN_IF_ERROR(
registry.UpdateCommandStats(metadata.id, executed_commands));
RETURN_IF_ERROR(registry.AppendLog(
metadata.id,
absl::StrCat("Commands executed: ", executed_commands)));
std::filesystem::path proposal_dir = metadata.log_path.parent_path();
std::filesystem::path proposal_path = proposal_dir / "proposal.json";
auto save_proposal_status =
generator.SaveProposal(proposal, proposal_path.string());
if (!save_proposal_status.ok()) {
@@ -198,16 +267,22 @@ absl::Status HandleRunCommand(const std::vector<std::string>& arg_vec,
save_proposal_status.message()));
}
RETURN_IF_ERROR(registry.AppendLog(
metadata.id,
absl::StrCat("Saved proposal JSON to ", proposal_path.string())));
std::cout
<< "✅ Agent successfully planned and executed changes in a sandbox."
<< std::endl;
std::cout << " Proposal ID: " << proposal.id << std::endl;
std::cout << " Proposal ID: " << metadata.id << std::endl;
std::cout << " Sandbox ROM: " << sandbox.rom_path << std::endl;
std::cout << " Proposal file: " << proposal_path << std::endl;
std::cout << " Proposal dir: " << proposal_dir << std::endl;
std::cout << " Diff file: " << metadata.diff_path << std::endl;
std::cout << " Log file: " << metadata.log_path << std::endl;
std::cout << "\nTo review the changes, run:\n";
std::cout << " z3ed agent diff --proposal-id " << proposal.id << std::endl;
std::cout << " z3ed agent diff --proposal-id " << metadata.id << std::endl;
std::cout << "\nTo accept the changes, run:\n";
std::cout << " z3ed agent accept --proposal-id " << proposal.id << std::endl;
std::cout << " z3ed agent accept --proposal-id " << metadata.id << std::endl;
return absl::OkStatus();
}
@@ -287,6 +362,14 @@ absl::Status HandleDiffCommand(Rom& rom, const std::vector<std::string>& args) {
std::cout << "Commands Executed: " << proposal.commands_executed << "\n";
std::cout << "Bytes Changed: " << proposal.bytes_changed << "\n\n";
if (!proposal.sandbox_rom_path.empty()) {
std::cout << "Sandbox ROM: " << proposal.sandbox_rom_path << "\n";
}
std::cout << "Proposal directory: "
<< proposal.log_path.parent_path() << "\n";
std::cout << "Diff file: " << proposal.diff_path << "\n";
std::cout << "Log file: " << proposal.log_path << "\n\n";
if (std::filesystem::exists(proposal.diff_path)) {
std::cout << "--- Diff Content ---\n";
std::ifstream diff_file(proposal.diff_path);
@@ -490,45 +573,90 @@ absl::Status HandleChatCommand(Rom& rom) {
absl::Status HandleAcceptCommand(const std::vector<std::string>& arg_vec,
Rom& rom) {
if (arg_vec.empty() || arg_vec[0] != "--proposal-id") {
std::optional<std::string> proposal_id;
for (size_t i = 0; i < arg_vec.size(); ++i) {
const std::string& token = arg_vec[i];
if (absl::StartsWith(token, "--proposal-id=")) {
proposal_id = token.substr(14);
break;
}
if (token == "--proposal-id" && i + 1 < arg_vec.size()) {
proposal_id = arg_vec[i + 1];
break;
}
}
if (!proposal_id.has_value() || proposal_id->empty()) {
return absl::InvalidArgumentError(
"Usage: agent accept --proposal-id <proposal_id>");
}
std::string proposal_id = arg_vec[1];
// 1. Load the proposal from disk.
Tile16ProposalGenerator generator;
auto proposal_path =
RomSandboxManager::Instance().RootDirectory() / (proposal_id + ".json");
auto proposal_or = generator.LoadProposal(proposal_path.string());
if (!proposal_or.ok()) {
return absl::InternalError(
absl::StrCat("Failed to load proposal file '", proposal_path.string(),
"': ", proposal_or.status().message()));
auto& registry = ProposalRegistry::Instance();
ASSIGN_OR_RETURN(auto metadata, registry.GetProposal(*proposal_id));
if (metadata.status == ProposalRegistry::ProposalStatus::kAccepted) {
std::cout << "Proposal '" << *proposal_id << "' is already accepted."
<< std::endl;
return absl::OkStatus();
}
auto proposal = proposal_or.value();
// 2. Ensure the main ROM is loaded.
RETURN_IF_ERROR(EnsureRomLoaded(rom, "agent accept --proposal-id <id>"));
if (metadata.sandbox_rom_path.empty()) {
return absl::FailedPreconditionError(absl::StrCat(
"Proposal '", *proposal_id,
"' is missing sandbox ROM metadata. Cannot accept."));
}
// 3. Apply the proposal to the main ROM.
auto apply_status = generator.ApplyProposal(proposal, &rom);
if (!apply_status.ok()) {
if (!std::filesystem::exists(metadata.sandbox_rom_path)) {
return absl::NotFoundError(absl::StrCat(
"Sandbox ROM not found at ", metadata.sandbox_rom_path.string()));
}
RETURN_IF_ERROR(
EnsureRomLoaded(rom, "agent accept --proposal-id <proposal_id>"));
Rom sandbox_rom;
auto sandbox_load_status = sandbox_rom.LoadFromFile(
metadata.sandbox_rom_path.string(), RomLoadOptions::CliDefaults());
if (!sandbox_load_status.ok()) {
return absl::InternalError(absl::StrCat(
"Failed to apply proposal to main ROM: ", apply_status.message()));
"Failed to load sandbox ROM: ", sandbox_load_status.message()));
}
if (rom.size() != sandbox_rom.size()) {
rom.Expand(static_cast<int>(sandbox_rom.size()));
}
auto copy_status = rom.WriteVector(0, sandbox_rom.vector());
if (!copy_status.ok()) {
return absl::InternalError(absl::StrCat(
"Failed to copy sandbox ROM data: ", copy_status.message()));
}
// 4. Save the changes to the main ROM file.
auto save_status = rom.SaveToFile({.save_new = false});
if (!save_status.ok()) {
return absl::InternalError(absl::StrCat(
"Failed to save changes to main ROM: ", save_status.message()));
}
std::cout << "✅ Proposal '" << proposal_id << "' accepted and applied to '"
<< rom.filename() << "'." << std::endl;
RETURN_IF_ERROR(registry.UpdateStatus(
*proposal_id, ProposalRegistry::ProposalStatus::kAccepted));
RETURN_IF_ERROR(registry.AppendLog(
*proposal_id,
absl::StrCat("Proposal accepted and applied to ", rom.filename())));
// TODO: Clean up sandbox and proposal files.
if (!metadata.sandbox_id.empty()) {
auto remove_status =
RomSandboxManager::Instance().RemoveSandbox(metadata.sandbox_id);
if (!remove_status.ok()) {
std::cerr << "Warning: Failed to remove sandbox '" << metadata.sandbox_id
<< "': " << remove_status.message() << "\n";
}
}
std::cout << "✅ Proposal '" << *proposal_id << "' accepted and applied to '"
<< rom.filename() << "'." << std::endl;
std::cout << " Source sandbox ROM: " << metadata.sandbox_rom_path
<< std::endl;
return absl::OkStatus();
}

View File

@@ -1,12 +1,21 @@
#include "cli/handlers/agent/commands.h"
#include <filesystem>
#include <fstream>
#include <iostream>
#include <optional>
#include <string>
#include <vector>
#include "absl/status/status.h"
#include "absl/status/statusor.h"
#include "absl/strings/numbers.h"
#include "absl/strings/str_cat.h"
#include "absl/strings/str_format.h"
#include "absl/time/time.h"
#include "cli/handlers/agent/common.h"
#include "nlohmann/json.hpp"
#include "util/macro.h"
#ifdef YAZE_WITH_GRPC
#include "cli/service/gui/gui_automation_client.h"
@@ -19,6 +28,100 @@ namespace agent {
#ifdef YAZE_WITH_GRPC
namespace {
struct RecordingState {
std::string recording_id;
std::string host = "localhost";
int port = 50052;
std::string output_path;
};
std::filesystem::path RecordingStateFilePath() {
std::error_code ec;
std::filesystem::path base =
std::filesystem::temp_directory_path(ec);
if (ec) {
base = std::filesystem::current_path();
}
return base / "yaze" / "agent" / "recording_state.json";
}
absl::Status SaveRecordingState(const RecordingState& state) {
auto path = RecordingStateFilePath();
std::error_code ec;
std::filesystem::create_directories(path.parent_path(), ec);
nlohmann::json json;
json["recording_id"] = state.recording_id;
json["host"] = state.host;
json["port"] = state.port;
json["output_path"] = state.output_path;
std::ofstream out(path, std::ios::out | std::ios::trunc);
if (!out.is_open()) {
return absl::InternalError(absl::StrCat("Failed to write recording state to ",
path.string()));
}
out << json.dump(2);
if (!out.good()) {
return absl::InternalError(
absl::StrCat("Failed to flush recording state to ", path.string()));
}
return absl::OkStatus();
}
absl::StatusOr<RecordingState> LoadRecordingState() {
auto path = RecordingStateFilePath();
std::ifstream in(path);
if (!in.is_open()) {
return absl::NotFoundError("No active recording session found. Run 'z3ed agent test record start' first.");
}
nlohmann::json json;
try {
in >> json;
} catch (const nlohmann::json::parse_error& error) {
return absl::InternalError(
absl::StrCat("Failed to parse recording state at ", path.string(),
": ", error.what()));
}
RecordingState state;
state.recording_id = json.value("recording_id", "");
state.host = json.value("host", "localhost");
state.port = json.value("port", 50052);
state.output_path = json.value("output_path", "");
if (state.recording_id.empty()) {
return absl::InvalidArgumentError(
absl::StrCat("Recording state at ", path.string(),
" is missing a recording_id"));
}
return state;
}
absl::Status ClearRecordingState() {
auto path = RecordingStateFilePath();
std::error_code ec;
std::filesystem::remove(path, ec);
if (ec && ec != std::errc::no_such_file_or_directory) {
return absl::InternalError(absl::StrCat("Failed to clear recording state: ",
ec.message()));
}
return absl::OkStatus();
}
std::string DefaultRecordingOutputPath() {
absl::Time now = absl::Now();
return absl::StrCat("tests/gui/recording-",
absl::FormatTime("%Y%m%dT%H%M%S", now,
absl::LocalTimeZone()),
".json");
}
} // namespace
// Forward declarations for subcommand handlers
absl::Status HandleTestRunCommand(const std::vector<std::string>& args);
absl::Status HandleTestReplayCommand(const std::vector<std::string>& args);
@@ -307,7 +410,10 @@ absl::Status HandleTestResultsCommand(const std::vector<std::string>& args) {
absl::Status HandleTestRecordCommand(const std::vector<std::string>& args) {
if (args.empty()) {
return absl::InvalidArgumentError(
"Usage: agent test record <start|stop> [--output <file>]");
"Usage: agent test record <start|stop> [options]\n"
" start [--output <file>] [--description <text>] [--session <id>]\n"
" [--host <host>] [--port <port>]\n"
" stop [--validate] [--discard] [--host <host>] [--port <port>]");
}
std::string action = args[0];
@@ -315,10 +421,167 @@ absl::Status HandleTestRecordCommand(const std::vector<std::string>& args) {
return absl::InvalidArgumentError("Record action must be 'start' or 'stop'");
}
// TODO: Implement recording functionality
return absl::UnimplementedError(
"Test recording is not yet implemented.\n"
"This feature will allow capturing GUI interactions for replay.");
if (action == "start") {
std::string host = "localhost";
int port = 50052;
std::string description;
std::string session_name;
std::string output_path;
for (size_t i = 1; i < args.size(); ++i) {
const std::string& token = args[i];
if (token == "--output" && i + 1 < args.size()) {
output_path = args[++i];
} else if (token == "--description" && i + 1 < args.size()) {
description = args[++i];
} else if (token == "--session" && i + 1 < args.size()) {
session_name = args[++i];
} else if (token == "--host" && i + 1 < args.size()) {
host = args[++i];
} else if (token == "--port" && i + 1 < args.size()) {
std::string port_value = args[++i];
int parsed_port = 0;
if (!absl::SimpleAtoi(port_value, &parsed_port)) {
return absl::InvalidArgumentError(
absl::StrCat("Invalid --port value: ", port_value));
}
port = parsed_port;
}
}
if (output_path.empty()) {
output_path = DefaultRecordingOutputPath();
}
std::filesystem::path absolute_output =
std::filesystem::absolute(output_path);
std::error_code ec;
std::filesystem::create_directories(absolute_output.parent_path(), ec);
GuiAutomationClient client(absl::StrCat(host, ":", port));
RETURN_IF_ERROR(client.Connect());
if (session_name.empty()) {
session_name = std::filesystem::path(output_path).stem().string();
}
ASSIGN_OR_RETURN(auto start_result,
client.StartRecording(absolute_output.string(),
session_name, description));
if (!start_result.success) {
return absl::InternalError(
absl::StrCat("Harness rejected start-recording request: ",
start_result.message));
}
RecordingState state;
state.recording_id = start_result.recording_id;
state.host = host;
state.port = port;
state.output_path = absolute_output.string();
RETURN_IF_ERROR(SaveRecordingState(state));
std::cout << "\n=== Recording Session Started ===\n";
std::cout << "Recording ID: " << start_result.recording_id << "\n";
std::cout << "Server: " << host << ":" << port << "\n";
std::cout << "Output: " << absolute_output << "\n";
if (!description.empty()) {
std::cout << "Description: " << description << "\n";
}
if (start_result.started_at.has_value()) {
std::cout << "Started: "
<< absl::FormatTime("%Y-%m-%d %H:%M:%S",
*start_result.started_at,
absl::LocalTimeZone())
<< "\n";
}
std::cout << "\nPress Ctrl+C to abort the recording session.\n";
return absl::OkStatus();
}
// Stop
bool validate = false;
bool discard = false;
std::optional<std::string> host_override;
std::optional<int> port_override;
for (size_t i = 1; i < args.size(); ++i) {
const std::string& token = args[i];
if (token == "--validate") {
validate = true;
} else if (token == "--discard") {
discard = true;
} else if (token == "--host" && i + 1 < args.size()) {
host_override = args[++i];
} else if (token == "--port" && i + 1 < args.size()) {
std::string port_value = args[++i];
int parsed_port = 0;
if (!absl::SimpleAtoi(port_value, &parsed_port)) {
return absl::InvalidArgumentError(
absl::StrCat("Invalid --port value: ", port_value));
}
port_override = parsed_port;
}
}
if (discard && validate) {
return absl::InvalidArgumentError(
"Cannot use --validate and --discard together");
}
ASSIGN_OR_RETURN(auto state, LoadRecordingState());
if (host_override.has_value()) {
state.host = *host_override;
}
if (port_override.has_value()) {
state.port = *port_override;
}
GuiAutomationClient client(absl::StrCat(state.host, ":", state.port));
RETURN_IF_ERROR(client.Connect());
ASSIGN_OR_RETURN(auto stop_result,
client.StopRecording(state.recording_id, discard));
if (!stop_result.success) {
return absl::InternalError(
absl::StrCat("Stop recording failed: ", stop_result.message));
}
RETURN_IF_ERROR(ClearRecordingState());
std::cout << "\n=== Recording Session Completed ===\n";
std::cout << "Recording ID: " << state.recording_id << "\n";
std::cout << "Server: " << state.host << ":" << state.port << "\n";
std::cout << "Steps captured: " << stop_result.step_count << "\n";
std::cout << "Duration: " << stop_result.duration.count() << " ms\n";
if (!stop_result.message.empty()) {
std::cout << "Message: " << stop_result.message << "\n";
}
if (!discard && !stop_result.output_path.empty()) {
std::cout << "Output saved to: " << stop_result.output_path << "\n";
}
if (discard) {
std::cout << "Recording discarded; no script file was produced." << std::endl;
return absl::OkStatus();
}
if (!validate || stop_result.output_path.empty()) {
std::cout << std::endl;
return absl::OkStatus();
}
std::cout << "\nReplaying recorded script to validate...\n";
ASSIGN_OR_RETURN(auto replay_result,
client.ReplayTest(stop_result.output_path, false, {}));
if (!replay_result.success) {
return absl::InternalError(
absl::StrCat("Replay failed: ", replay_result.message));
}
std::cout << "Replay succeeded. Steps executed: "
<< replay_result.steps_executed << "\n";
return absl::OkStatus();
}
#endif // YAZE_WITH_GRPC

View File

@@ -186,6 +186,74 @@ absl::StatusOr<ReplayTestResult> GuiAutomationClient::ReplayTest(
#endif
}
absl::StatusOr<StartRecordingResult> GuiAutomationClient::StartRecording(
const std::string& output_path, const std::string& session_name,
const std::string& description) {
#ifdef YAZE_WITH_GRPC
if (!stub_) {
return absl::FailedPreconditionError("Not connected. Call Connect() first.");
}
yaze::test::StartRecordingRequest request;
request.set_output_path(output_path);
request.set_session_name(session_name);
request.set_description(description);
yaze::test::StartRecordingResponse response;
grpc::ClientContext context;
grpc::Status status = stub_->StartRecording(&context, request, &response);
if (!status.ok()) {
return absl::InternalError(
absl::StrCat("StartRecording RPC failed: ", status.error_message()));
}
StartRecordingResult result;
result.success = response.success();
result.message = response.message();
result.recording_id = response.recording_id();
result.started_at = OptionalTimeFromMillis(response.started_at_ms());
return result;
#else
return absl::UnimplementedError("gRPC not available");
#endif
}
absl::StatusOr<StopRecordingResult> GuiAutomationClient::StopRecording(
const std::string& recording_id, bool discard) {
#ifdef YAZE_WITH_GRPC
if (!stub_) {
return absl::FailedPreconditionError("Not connected. Call Connect() first.");
}
if (recording_id.empty()) {
return absl::InvalidArgumentError("recording_id must not be empty");
}
yaze::test::StopRecordingRequest request;
request.set_recording_id(recording_id);
request.set_discard(discard);
yaze::test::StopRecordingResponse response;
grpc::ClientContext context;
grpc::Status status = stub_->StopRecording(&context, request, &response);
if (!status.ok()) {
return absl::InternalError(
absl::StrCat("StopRecording RPC failed: ", status.error_message()));
}
StopRecordingResult result;
result.success = response.success();
result.message = response.message();
result.output_path = response.output_path();
result.step_count = response.step_count();
result.duration = std::chrono::milliseconds(response.duration_ms());
return result;
#else
return absl::UnimplementedError("gRPC not available");
#endif
}
absl::StatusOr<AutomationResult> GuiAutomationClient::Click(
const std::string& target, ClickType type) {
#ifdef YAZE_WITH_GRPC

View File

@@ -134,6 +134,21 @@ struct ReplayTestResult {
std::vector<std::string> logs;
};
struct StartRecordingResult {
bool success = false;
std::string message;
std::string recording_id;
std::optional<absl::Time> started_at;
};
struct StopRecordingResult {
bool success = false;
std::string message;
std::string output_path;
int step_count = 0;
std::chrono::milliseconds duration{0};
};
enum class WidgetTypeFilter {
kUnspecified,
kAll,
@@ -303,6 +318,15 @@ class GuiAutomationClient {
const std::string& script_path, bool ci_mode,
const std::map<std::string, std::string>& parameter_overrides = {});
absl::StatusOr<StartRecordingResult> StartRecording(
const std::string& output_path,
const std::string& session_name,
const std::string& description);
absl::StatusOr<StopRecordingResult> StopRecording(
const std::string& recording_id,
bool discard = false);
/**
* @brief Check if client is connected
*/

View File

@@ -8,10 +8,13 @@
#include "absl/status/status.h"
#include "absl/status/statusor.h"
#include "absl/strings/ascii.h"
#include "absl/strings/match.h"
#include "absl/strings/str_cat.h"
#include "absl/strings/str_format.h"
#include "absl/time/time.h"
#include "nlohmann/json.hpp"
#include "util/macro.h"
namespace yaze {
@@ -31,6 +34,40 @@ std::filesystem::path DetermineDefaultRoot() {
return temp_dir / "yaze" / "proposals";
}
std::string StatusToString(ProposalRegistry::ProposalStatus status) {
switch (status) {
case ProposalRegistry::ProposalStatus::kAccepted:
return "accepted";
case ProposalRegistry::ProposalStatus::kRejected:
return "rejected";
case ProposalRegistry::ProposalStatus::kPending:
default:
return "pending";
}
}
ProposalRegistry::ProposalStatus ParseStatus(absl::string_view value) {
std::string lower = absl::AsciiStrToLower(value);
if (absl::StartsWith(lower, "accept")) {
return ProposalRegistry::ProposalStatus::kAccepted;
}
if (absl::StartsWith(lower, "reject")) {
return ProposalRegistry::ProposalStatus::kRejected;
}
return ProposalRegistry::ProposalStatus::kPending;
}
int64_t TimeToMillis(absl::Time time) {
return absl::ToUnixMillis(time);
}
std::optional<absl::Time> OptionalTimeFromMillis(int64_t millis) {
if (millis <= 0) {
return std::nullopt;
}
return absl::FromUnixMillis(millis);
}
} // namespace
ProposalRegistry& ProposalRegistry::Instance() {
@@ -65,92 +102,169 @@ absl::Status ProposalRegistry::EnsureRootExistsLocked() {
absl::Status ProposalRegistry::LoadProposalsFromDiskLocked() {
std::error_code ec;
// Check if root directory exists
if (!std::filesystem::exists(root_directory_, ec)) {
return absl::OkStatus(); // No proposals to load
return absl::OkStatus();
}
// Iterate over all directories in the root
for (const auto& entry : std::filesystem::directory_iterator(root_directory_, ec)) {
if (ec) {
continue; // Skip entries that cause errors
}
if (!entry.is_directory()) {
continue; // Skip non-directories
break;
}
std::string proposal_id = entry.path().filename().string();
// Skip if already loaded (shouldn't happen, but be defensive)
if (!entry.is_directory()) {
continue;
}
const std::string proposal_id = entry.path().filename().string();
if (proposals_.find(proposal_id) != proposals_.end()) {
continue;
}
// Reconstruct metadata from directory contents
// Since we don't have a metadata.json file, we need to infer what we can
std::filesystem::path log_path = entry.path() / "execution.log";
std::filesystem::path diff_path = entry.path() / "diff.txt";
// Check if log file exists to determine if this is a valid proposal
if (!std::filesystem::exists(log_path, ec)) {
continue; // Not a valid proposal directory
}
ProposalMetadata metadata;
bool metadata_loaded = false;
const std::filesystem::path metadata_path = entry.path() / "metadata.json";
// Extract timestamp from proposal ID (format: proposal-20251001T200215-1)
absl::Time created_at = absl::Now(); // Default to now if parsing fails
if (proposal_id.starts_with("proposal-")) {
std::string time_str = proposal_id.substr(9, 15); // Extract YYYYMMDDTHHmmSS
std::string error;
if (absl::ParseTime("%Y%m%dT%H%M%S", time_str, &created_at, &error)) {
// Successfully parsed time
if (std::filesystem::exists(metadata_path, ec) && !ec) {
std::ifstream metadata_file(metadata_path);
if (metadata_file.is_open()) {
try {
nlohmann::json metadata_json;
metadata_file >> metadata_json;
metadata.id = metadata_json.value("id", proposal_id);
if (metadata.id.empty()) {
metadata.id = proposal_id;
}
metadata.sandbox_id = metadata_json.value("sandbox_id", "");
if (metadata_json.contains("sandbox_directory") &&
metadata_json["sandbox_directory"].is_string()) {
metadata.sandbox_directory =
std::filesystem::path(metadata_json["sandbox_directory"].get<std::string>());
} else {
metadata.sandbox_directory.clear();
}
if (metadata_json.contains("sandbox_rom_path") &&
metadata_json["sandbox_rom_path"].is_string()) {
metadata.sandbox_rom_path =
std::filesystem::path(metadata_json["sandbox_rom_path"].get<std::string>());
} else {
metadata.sandbox_rom_path.clear();
}
metadata.description = metadata_json.value("description", "");
metadata.prompt = metadata_json.value("prompt", "");
metadata.status = ParseStatus(metadata_json.value("status", "pending"));
int64_t created_at_millis = metadata_json.value<int64_t>(
"created_at_millis", TimeToMillis(absl::Now()));
metadata.created_at = absl::FromUnixMillis(created_at_millis);
int64_t reviewed_at_millis = metadata_json.value<int64_t>(
"reviewed_at_millis", 0);
metadata.reviewed_at = OptionalTimeFromMillis(reviewed_at_millis);
std::string diff_path = metadata_json.value("diff_path", std::string("diff.txt"));
std::string log_path = metadata_json.value("log_path", std::string("execution.log"));
metadata.diff_path = entry.path() / diff_path;
metadata.log_path = entry.path() / log_path;
metadata.bytes_changed = metadata_json.value("bytes_changed", 0);
metadata.commands_executed = metadata_json.value("commands_executed", 0);
metadata.screenshots.clear();
if (metadata_json.contains("screenshots") &&
metadata_json["screenshots"].is_array()) {
for (const auto& screenshot : metadata_json["screenshots"]) {
if (screenshot.is_string()) {
metadata.screenshots.emplace_back(entry.path() /
screenshot.get<std::string>());
}
}
}
if (metadata.sandbox_directory.empty() &&
!metadata.sandbox_rom_path.empty()) {
metadata.sandbox_directory = metadata.sandbox_rom_path.parent_path();
}
metadata_loaded = true;
} catch (const std::exception& ex) {
std::cerr << "Warning: Failed to parse metadata for proposal "
<< proposal_id << ": " << ex.what() << "\n";
}
}
}
// Get file modification time as a fallback
auto ftime = std::filesystem::last_write_time(log_path, ec);
if (!ec) {
auto sctp = std::chrono::time_point_cast<std::chrono::system_clock::duration>(
ftime - std::filesystem::file_time_type::clock::now() +
std::chrono::system_clock::now());
auto time_t_value = std::chrono::system_clock::to_time_t(sctp);
created_at = absl::FromTimeT(time_t_value);
}
if (!metadata_loaded) {
std::filesystem::path log_path = entry.path() / "execution.log";
if (!std::filesystem::exists(log_path, ec) || ec) {
continue;
}
// Create minimal metadata for this proposal
ProposalMetadata metadata{
.id = proposal_id,
.sandbox_id = "", // Unknown - not stored in logs
.description = "Loaded from disk",
.prompt = "", // Unknown - not stored in logs
.status = ProposalStatus::kPending,
.created_at = created_at,
.reviewed_at = std::nullopt,
.diff_path = diff_path,
.log_path = log_path,
.screenshots = {},
.bytes_changed = 0,
.commands_executed = 0,
};
std::filesystem::path diff_path = entry.path() / "diff.txt";
// Count diff size if it exists
if (std::filesystem::exists(diff_path, ec) && !ec) {
metadata.bytes_changed = static_cast<int>(
std::filesystem::file_size(diff_path, ec));
}
absl::Time created_at = absl::Now();
if (proposal_id.starts_with("proposal-")) {
std::string time_str = proposal_id.substr(9, 15);
std::string error;
if (absl::ParseTime("%Y%m%dT%H%M%S", time_str, &created_at, &error)) {
// Parsed successfully.
}
}
// Scan for screenshots
for (const auto& file : std::filesystem::directory_iterator(entry.path(), ec)) {
if (ec) continue;
if (file.path().extension() == ".png" ||
file.path().extension() == ".jpg" ||
file.path().extension() == ".jpeg") {
metadata.screenshots.push_back(file.path());
auto ftime = std::filesystem::last_write_time(log_path, ec);
if (!ec) {
auto sctp = std::chrono::time_point_cast<std::chrono::system_clock::duration>(
ftime - std::filesystem::file_time_type::clock::now() +
std::chrono::system_clock::now());
auto time_t_value = std::chrono::system_clock::to_time_t(sctp);
created_at = absl::FromTimeT(time_t_value);
}
metadata = ProposalMetadata{
.id = proposal_id,
.sandbox_id = "",
.sandbox_directory = std::filesystem::path(),
.sandbox_rom_path = std::filesystem::path(),
.description = "Loaded from disk",
.prompt = "",
.status = ProposalStatus::kPending,
.created_at = created_at,
.reviewed_at = std::nullopt,
.diff_path = diff_path,
.log_path = log_path,
.screenshots = {},
.bytes_changed = 0,
.commands_executed = 0,
};
if (std::filesystem::exists(diff_path, ec) && !ec) {
metadata.bytes_changed = static_cast<int>(
std::filesystem::file_size(diff_path, ec));
}
for (const auto& file : std::filesystem::directory_iterator(entry.path(), ec)) {
if (ec) {
break;
}
if (file.path().extension() == ".png" || file.path().extension() == ".jpg" ||
file.path().extension() == ".jpeg") {
metadata.screenshots.push_back(file.path());
}
}
// Create a metadata file for legacy proposals so future loads are fast.
absl::Status write_status = WriteMetadataLocked(metadata);
if (!write_status.ok()) {
std::cerr << "Warning: Failed to persist metadata for legacy proposal "
<< proposal_id << ": " << write_status.message() << "\n";
}
}
proposals_[proposal_id] = metadata;
proposals_[metadata.id] = metadata;
}
return absl::OkStatus();
@@ -171,6 +285,7 @@ std::filesystem::path ProposalRegistry::ProposalDirectory(
absl::StatusOr<ProposalRegistry::ProposalMetadata>
ProposalRegistry::CreateProposal(absl::string_view sandbox_id,
const std::filesystem::path& sandbox_rom_path,
absl::string_view prompt,
absl::string_view description) {
std::unique_lock<std::mutex> lock(mutex_);
@@ -191,6 +306,10 @@ ProposalRegistry::CreateProposal(absl::string_view sandbox_id,
proposals_[id] = ProposalMetadata{
.id = id,
.sandbox_id = std::string(sandbox_id),
.sandbox_directory = sandbox_rom_path.empty()
? std::filesystem::path()
: sandbox_rom_path.parent_path(),
.sandbox_rom_path = sandbox_rom_path,
.description = std::string(description),
.prompt = std::string(prompt),
.status = ProposalStatus::kPending,
@@ -203,6 +322,8 @@ ProposalRegistry::CreateProposal(absl::string_view sandbox_id,
.commands_executed = 0,
};
RETURN_IF_ERROR(WriteMetadataLocked(proposals_.at(id)));
return proposals_.at(id);
}
@@ -227,6 +348,8 @@ absl::Status ProposalRegistry::RecordDiff(const std::string& proposal_id,
// Update bytes_changed metric (rough estimate based on diff size)
it->second.bytes_changed = static_cast<int>(diff_content.size());
RETURN_IF_ERROR(WriteMetadataLocked(it->second));
return absl::OkStatus();
}
@@ -272,6 +395,21 @@ absl::Status ProposalRegistry::AddScreenshot(
}
it->second.screenshots.push_back(screenshot_path);
RETURN_IF_ERROR(WriteMetadataLocked(it->second));
return absl::OkStatus();
}
absl::Status ProposalRegistry::UpdateCommandStats(const std::string& proposal_id,
int commands_executed) {
std::lock_guard<std::mutex> lock(mutex_);
auto it = proposals_.find(proposal_id);
if (it == proposals_.end()) {
return absl::NotFoundError(
absl::StrCat("Proposal not found: ", proposal_id));
}
it->second.commands_executed = commands_executed;
RETURN_IF_ERROR(WriteMetadataLocked(it->second));
return absl::OkStatus();
}
@@ -286,7 +424,7 @@ absl::Status ProposalRegistry::UpdateStatus(const std::string& proposal_id,
it->second.status = status;
it->second.reviewed_at = absl::Now();
RETURN_IF_ERROR(WriteMetadataLocked(it->second));
return absl::OkStatus();
}
@@ -354,6 +492,70 @@ ProposalRegistry::GetLatestPendingProposal() const {
return *latest;
}
absl::Status ProposalRegistry::WriteMetadataLocked(
const ProposalMetadata& metadata) const {
std::filesystem::path proposal_dir = ProposalDirectory(metadata.id);
std::error_code ec;
if (!std::filesystem::exists(proposal_dir, ec) || ec) {
return absl::NotFoundError(
absl::StrCat("Proposal directory missing for ", metadata.id));
}
auto relative_to_proposal = [&](const std::filesystem::path& path) {
if (path.empty()) {
return std::string();
}
std::error_code relative_error;
auto relative_path = std::filesystem::relative(path, proposal_dir, relative_error);
if (!relative_error) {
return relative_path.generic_string();
}
return path.generic_string();
};
nlohmann::json metadata_json;
metadata_json["id"] = metadata.id;
metadata_json["sandbox_id"] = metadata.sandbox_id;
if (!metadata.sandbox_directory.empty()) {
metadata_json["sandbox_directory"] = metadata.sandbox_directory.generic_string();
}
if (!metadata.sandbox_rom_path.empty()) {
metadata_json["sandbox_rom_path"] = metadata.sandbox_rom_path.generic_string();
}
metadata_json["description"] = metadata.description;
metadata_json["prompt"] = metadata.prompt;
metadata_json["status"] = StatusToString(metadata.status);
metadata_json["created_at_millis"] = TimeToMillis(metadata.created_at);
metadata_json["reviewed_at_millis"] = metadata.reviewed_at.has_value()
? TimeToMillis(*metadata.reviewed_at)
: int64_t{0};
metadata_json["diff_path"] = relative_to_proposal(metadata.diff_path);
metadata_json["log_path"] = relative_to_proposal(metadata.log_path);
metadata_json["bytes_changed"] = metadata.bytes_changed;
metadata_json["commands_executed"] = metadata.commands_executed;
nlohmann::json screenshots_json = nlohmann::json::array();
for (const auto& screenshot : metadata.screenshots) {
screenshots_json.push_back(relative_to_proposal(screenshot));
}
metadata_json["screenshots"] = std::move(screenshots_json);
std::ofstream metadata_file(proposal_dir / "metadata.json", std::ios::out);
if (!metadata_file.is_open()) {
return absl::InternalError(absl::StrCat(
"Failed to write metadata file for proposal ", metadata.id));
}
metadata_file << metadata_json.dump(2);
metadata_file.close();
if (!metadata_file) {
return absl::InternalError(absl::StrCat(
"Failed to flush metadata file for proposal ", metadata.id));
}
return absl::OkStatus();
}
absl::Status ProposalRegistry::RemoveProposal(const std::string& proposal_id) {
std::lock_guard<std::mutex> lock(mutex_);
auto it = proposals_.find(proposal_id);

View File

@@ -37,6 +37,8 @@ class ProposalRegistry {
struct ProposalMetadata {
std::string id;
std::string sandbox_id;
std::filesystem::path sandbox_directory;
std::filesystem::path sandbox_rom_path;
std::string description;
std::string prompt; // Original agent prompt that created this proposal
ProposalStatus status;
@@ -65,6 +67,7 @@ class ProposalRegistry {
// is created under the root, and metadata is initialized.
absl::StatusOr<ProposalMetadata> CreateProposal(
absl::string_view sandbox_id,
const std::filesystem::path& sandbox_rom_path,
absl::string_view prompt,
absl::string_view description);
@@ -82,6 +85,11 @@ class ProposalRegistry {
absl::Status AddScreenshot(const std::string& proposal_id,
const std::filesystem::path& screenshot_path);
// Updates the number of commands executed for a proposal. Used to track
// how many CLI commands ran when generating the proposal.
absl::Status UpdateCommandStats(const std::string& proposal_id,
int commands_executed);
// Updates the proposal status (pending -> accepted/rejected) and sets
// the review timestamp.
absl::Status UpdateStatus(const std::string& proposal_id,
@@ -111,6 +119,7 @@ class ProposalRegistry {
absl::Status LoadProposalsFromDiskLocked();
std::string GenerateProposalIdLocked();
std::filesystem::path ProposalDirectory(absl::string_view proposal_id) const;
absl::Status WriteMetadataLocked(const ProposalMetadata& metadata) const;
std::filesystem::path root_directory_;
mutable std::mutex mutex_;

View File

@@ -1,12 +1,15 @@
#include "cli/service/planning/tile16_proposal_generator.h"
#include <sstream>
#include <fstream>
#include <sstream>
#include "absl/strings/match.h"
#include "absl/strings/str_split.h"
#include "absl/strings/str_cat.h"
#include "absl/strings/numbers.h"
#include "app/zelda3/overworld/overworld.h"
#include "nlohmann/json.hpp"
#include "util/macro.h"
namespace yaze {
namespace cli {
@@ -54,9 +57,175 @@ std::string Tile16Proposal::ToJson() const {
return json.str();
}
absl::StatusOr<Tile16Proposal> Tile16Proposal::FromJson(const std::string& /* json */) {
// TODO: Implement JSON parsing using nlohmann/json when available
return absl::UnimplementedError("JSON parsing not yet implemented");
namespace {
absl::StatusOr<uint16_t> ParseTileValue(const nlohmann::json& json,
const char* field) {
if (!json.contains(field)) {
return absl::InvalidArgumentError(
absl::StrCat("Missing field '", field, "' in proposal change"));
}
if (json[field].is_number_integer()) {
int value = json[field].get<int>();
if (value < 0 || value > 0xFFFF) {
return absl::InvalidArgumentError(
absl::StrCat("Tile value for '", field,
"' out of range: ", value));
}
return static_cast<uint16_t>(value);
}
if (json[field].is_string()) {
std::string value = json[field].get<std::string>();
if (value.empty()) {
return absl::InvalidArgumentError(
absl::StrCat("Tile value for '", field, "' is empty"));
}
// Support hex strings in 0xFFFF format or plain decimal strings
if (absl::StartsWith(value, "0x") || absl::StartsWith(value, "0X")) {
if (value.size() <= 2) {
return absl::InvalidArgumentError(
absl::StrCat("Invalid hex tile value for '", field,
"': ", json[field].get<std::string>()));
}
value = value.substr(2);
unsigned int parsed = 0;
if (!absl::SimpleHexAtoi(value, &parsed) || parsed > 0xFFFF) {
return absl::InvalidArgumentError(
absl::StrCat("Invalid hex tile value for '", field,
"': ", json[field].get<std::string>()));
}
return static_cast<uint16_t>(parsed);
}
unsigned int parsed = 0;
if (!absl::SimpleAtoi(value, &parsed) || parsed > 0xFFFF) {
return absl::InvalidArgumentError(
absl::StrCat("Invalid tile value for '", field,
"': ", json[field].get<std::string>()));
}
return static_cast<uint16_t>(parsed);
}
return absl::InvalidArgumentError(
absl::StrCat("Unsupported JSON type for tile field '", field, "'"));
}
Tile16Proposal::Status ParseStatus(absl::string_view status_text) {
if (absl::StartsWith(status_text, "accept")) {
return Tile16Proposal::Status::ACCEPTED;
}
if (absl::StartsWith(status_text, "reject")) {
return Tile16Proposal::Status::REJECTED;
}
if (absl::StartsWith(status_text, "apply")) {
return Tile16Proposal::Status::APPLIED;
}
return Tile16Proposal::Status::PENDING;
}
} // namespace
absl::StatusOr<Tile16Proposal> Tile16Proposal::FromJson(
const std::string& json_text) {
nlohmann::json json;
try {
json = nlohmann::json::parse(json_text);
} catch (const nlohmann::json::parse_error& error) {
return absl::InvalidArgumentError(
absl::StrCat("Failed to parse proposal JSON: ", error.what()));
}
Tile16Proposal proposal;
if (!json.contains("id") || !json["id"].is_string()) {
return absl::InvalidArgumentError(
"Proposal JSON must include string field 'id'");
}
proposal.id = json["id"].get<std::string>();
if (!json.contains("prompt") || !json["prompt"].is_string()) {
return absl::InvalidArgumentError(
"Proposal JSON must include string field 'prompt'");
}
proposal.prompt = json["prompt"].get<std::string>();
if (json.contains("ai_service") && json["ai_service"].is_string()) {
proposal.ai_service = json["ai_service"].get<std::string>();
}
if (json.contains("reasoning") && json["reasoning"].is_string()) {
proposal.reasoning = json["reasoning"].get<std::string>();
}
if (json.contains("status")) {
if (!json["status"].is_string()) {
return absl::InvalidArgumentError(
"Proposal 'status' must be a string value");
}
proposal.status = ParseStatus(json["status"].get<std::string>());
} else {
proposal.status = Status::PENDING;
}
if (json.contains("changes")) {
if (!json["changes"].is_array()) {
return absl::InvalidArgumentError(
"Proposal 'changes' field must be an array");
}
for (const auto& change_json : json["changes"]) {
if (!change_json.is_object()) {
return absl::InvalidArgumentError(
"Each change entry must be a JSON object");
}
Tile16Change change;
if (!change_json.contains("map_id") ||
!change_json["map_id"].is_number_integer()) {
return absl::InvalidArgumentError(
"Tile change missing integer field 'map_id'");
}
change.map_id = change_json["map_id"].get<int>();
if (!change_json.contains("x") ||
!change_json["x"].is_number_integer()) {
return absl::InvalidArgumentError(
"Tile change missing integer field 'x'");
}
change.x = change_json["x"].get<int>();
if (!change_json.contains("y") ||
!change_json["y"].is_number_integer()) {
return absl::InvalidArgumentError(
"Tile change missing integer field 'y'");
}
change.y = change_json["y"].get<int>();
ASSIGN_OR_RETURN(change.old_tile,
ParseTileValue(change_json, "old_tile"));
ASSIGN_OR_RETURN(change.new_tile,
ParseTileValue(change_json, "new_tile"));
proposal.changes.push_back(change);
}
}
if (proposal.changes.empty()) {
return absl::InvalidArgumentError(
"Proposal JSON did not include any tile16 changes");
}
proposal.created_at = std::chrono::system_clock::now();
if (json.contains("created_at_ms") && json["created_at_ms"].is_number()) {
auto millis = json["created_at_ms"].get<int64_t>();
proposal.created_at = std::chrono::system_clock::time_point(
std::chrono::milliseconds(millis));
}
return proposal;
}
std::string Tile16ProposalGenerator::GenerateProposalId() const {

View File

@@ -267,7 +267,7 @@ endif()
target_link_libraries(yaze_test yaze_agent)
endif()
if(YAZE_BUILD_EMU AND NOT YAZE_WITH_GRPC AND TARGET yaze_emulator)
if(YAZE_BUILD_EMU AND TARGET yaze_emulator)
target_link_libraries(yaze_test yaze_emulator)
endif()
else()