diff --git a/doc/AGENT-LIFECYCLE.md b/doc/AGENT-LIFECYCLE.md index 96767c83131..87e0967316f 100644 --- a/doc/AGENT-LIFECYCLE.md +++ b/doc/AGENT-LIFECYCLE.md @@ -54,6 +54,19 @@ An old result cannot complete a newer operation. Deletion requires completed termination or a rejected hire. The existing restrictions for built-in agents and accounting still apply. +Messages accepted during setup remain queued until verification completes. +After the worker commits `ready`, its driver immediately asks the scheduler to +dispatch that agent's saved work. The scheduler rechecks company scope, +invokability, budgets, concurrency, and task ownership through normal admission. +If dispatch fails, readiness remains committed and the periodic queue sweep +retries; setup is not repeated because of a scheduler failure. + +Native OpenCode verification checks the selected model with the real hello +probe. It validates the model identifier but does not first enumerate the full model +catalog. Authentication failures, unavailable models, and probe timeouts still +prevent readiness. Without an explicit local working directory, the native +probe uses a temporary empty directory instead of the server's checkout. + ## Commands and transactions `requestHire` writes the agent record before preparation starts. diff --git a/packages/adapters/opencode-local/src/server/test.remote.test.ts b/packages/adapters/opencode-local/src/server/test.remote.test.ts index aebe6cfa0a9..b9a63a48b97 100644 --- a/packages/adapters/opencode-local/src/server/test.remote.test.ts +++ b/packages/adapters/opencode-local/src/server/test.remote.test.ts @@ -1,4 +1,4 @@ -import { mkdtemp, rm } from "node:fs/promises"; +import { mkdtemp, rm, stat } from "node:fs/promises"; import os from "node:os"; import path from "node:path"; import { afterEach, beforeEach, describe, expect, it, vi } from "vitest"; @@ -72,6 +72,13 @@ vi.mock("@paperclipai/adapter-utils/execution-target", async () => { }); import { testEnvironment } from "./test.js"; +import { discoverOpenCodeModels, ensureOpenCodeModelConfiguredAndAvailable } from "./models.js"; + +vi.mock("./models.js", async () => ({ + ...await vi.importActual("./models.js"), + discoverOpenCodeModels: vi.fn().mockRejectedValue(new Error("catalog should not be needed")), + ensureOpenCodeModelConfiguredAndAvailable: vi.fn().mockRejectedValue(new Error("catalog should not be needed")), +})); describe("opencode remote environment diagnostics", () => { const configHomes: string[] = []; @@ -89,6 +96,56 @@ describe("opencode remote environment diagnostics", () => { await Promise.all(configHomes.splice(0).map(dir => rm(dir, { recursive: true, force: true }))); }); + it.each(["passed", "auth_required", "model_unavailable", "timed_out"])( + "validates a native model with the real probe path and no catalog scan (%s)", async outcome => { + const successfulProbe = await runAdapterExecutionTargetProcess.getMockImplementation()!(); + runAdapterExecutionTargetProcess.mockResolvedValueOnce(outcome === "passed" ? successfulProbe : { + ...successfulProbe, exitCode: 1, stdout: "", timedOut: outcome === "timed_out", + stderr: outcome === "auth_required" ? "invalid API key" : "ProviderModelNotFoundError", + }); + const result = await testEnvironment({ companyId: "company-1", adapterType: "paperclip_runner", + config: { model: "openrouter/deepseek/deepseek-v4-flash-0731", env: { XDG_CONFIG_HOME: configHome } } }); + expect(discoverOpenCodeModels).not.toHaveBeenCalled(); + expect(ensureOpenCodeModelConfiguredAndAvailable).not.toHaveBeenCalled(); + expect(runAdapterExecutionTargetProcess).toHaveBeenCalledExactlyOnceWith( + expect.any(String), null, "opencode", ["run", "--format", "json", "--model", "openrouter/deepseek/deepseek-v4-flash-0731"], + expect.objectContaining({ stdin: "Respond with hello." }), + ); + expect(result.status).toBe(outcome === "passed" ? "pass" : "warn"); + expect(result.checks).toContainEqual(expect.objectContaining({ code: `opencode_hello_probe_${outcome}` })); + const call = runAdapterExecutionTargetProcess.mock.calls[0] as unknown as [unknown, unknown, unknown, unknown, { cwd: string }]; + expect(path.basename(call[4].cwd)).toMatch(/^paperclip-opencode-native-probe-/); + await expect(stat(call[4].cwd)).rejects.toMatchObject({ code: "ENOENT" }); + }, + ); + + it("rejects a malformed native model before starting the probe", async () => { + const result = await testEnvironment({ companyId: "company-1", adapterType: "paperclip_runner", + config: { model: "missing-provider", env: { XDG_CONFIG_HOME: configHome } } }); + expect(result.status).toBe("fail"); + expect(runAdapterExecutionTargetProcess).not.toHaveBeenCalled(); + expect(discoverOpenCodeModels).not.toHaveBeenCalled(); + }); + + it("preserves an explicitly configured native probe directory", async () => { + await testEnvironment({ companyId: "company-1", adapterType: "paperclip_runner", + config: { model: "provider/model", cwd: configHome, env: { XDG_CONFIG_HOME: configHome } } }); + const call = runAdapterExecutionTargetProcess.mock.calls[0] as unknown as [unknown, unknown, unknown, unknown, { cwd: string }]; + expect(call[4].cwd).toBe(configHome); + expect((await stat(configHome)).isDirectory()).toBe(true); + }); + + it("keeps catalog validation for the legacy local adapter", async () => { + vi.mocked(discoverOpenCodeModels).mockResolvedValueOnce([{ id: "provider/model", label: "Model" }]); + vi.mocked(ensureOpenCodeModelConfiguredAndAvailable).mockResolvedValueOnce([{ id: "provider/model", label: "Model" }]); + const result = await testEnvironment({ companyId: "company-1", adapterType: "opencode_local", + config: { model: "provider/model", env: { XDG_CONFIG_HOME: configHome } } }); + expect(result.status).toBe("pass"); + expect(discoverOpenCodeModels).toHaveBeenCalledOnce(); + expect(ensureOpenCodeModelConfiguredAndAvailable).toHaveBeenCalledOnce(); + expect(runAdapterExecutionTargetProcess).toHaveBeenCalledOnce(); + }); + it.each([false, true])("stages remote runtime config assets for sandbox hello probes (managed=%s)", async (managed) => { const configHome = await mkdtemp(path.join(os.tmpdir(), "opencode-remote-test-config-")); configHomes.push(configHome); diff --git a/packages/adapters/opencode-local/src/server/test.ts b/packages/adapters/opencode-local/src/server/test.ts index 7f2480b9c90..628745e9772 100644 --- a/packages/adapters/opencode-local/src/server/test.ts +++ b/packages/adapters/opencode-local/src/server/test.ts @@ -25,7 +25,7 @@ import { prepareAdapterExecutionTargetRuntime, overrideAdapterExecutionTargetRemoteCwd, } from "@paperclipai/adapter-utils/execution-target"; -import { discoverOpenCodeModels, ensureOpenCodeModelConfiguredAndAvailable } from "./models.js"; +import { discoverOpenCodeModels, ensureOpenCodeModelConfiguredAndAvailable, requireOpenCodeModelId } from "./models.js"; import { parseOpenCodeJsonl } from "./parse.js"; import { SANDBOX_INSTALL_COMMAND } from "../index.js"; import { prepareOpenCodeRuntimeConfig, prepareManagedOpenCodeRemoteHomes } from "./runtime-config.js"; @@ -141,9 +141,16 @@ export async function testEnvironment( // still has the path available for cleanup in `finally` — otherwise the // `fs.mkdtemp` directory leaks on the early-throw path. let preparedRuntimeWorkspaceLocalDir: string | null = null; + let nativeProbeWorkspaceLocalDir: string | null = null; try { let runtimeTarget: AdapterExecutionTarget | null = target ?? null; let runtimeCwd = cwd; + // Native workspaces are assigned at execution time. A configuration probe + // without an explicit workspace must not run in the server's checkout. + if (!targetIsRemote && ctx.adapterType === "paperclip_runner" && !asString(config.cwd, "").trim()) { + nativeProbeWorkspaceLocalDir = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-opencode-native-probe-")); + runtimeCwd = nativeProbeWorkspaceLocalDir; + } if (targetIsRemote) { preparedRuntimeWorkspaceLocalDir = await fs.mkdtemp(path.join(os.tmpdir(), `paperclip-opencode-envtest-${runId}-`)); const preparedExecutionTargetRuntime = await prepareAdapterExecutionTargetRuntime({ @@ -223,12 +230,30 @@ export async function testEnvironment( let modelValidationPassed = false; const configuredModel = asString(config.model, "").trim(); + // The native runner already selects a concrete model. Its real hello probe + // verifies model access and credentials; a full catalog scan adds another + // cold CLI launch (and discovery retries) without verifying that request. + const nativeModelProbe = ctx.adapterType === "paperclip_runner" && Boolean(configuredModel); + // Model discovery and validation use local child processes against // OpenCode's `models` subcommand and JSON config; these are not yet // wired through the execution target. When probing a remote env, skip // discovery/validation and rely on the remote hello probe to surface // model/auth issues directly. - if (targetIsRemote && configuredModel) { + if (nativeModelProbe) { + try { + requireOpenCodeModelId(configuredModel); + modelValidationPassed = true; + checks.push({ + code: "opencode_model_validation_via_probe", + level: "info", + message: "The configured native-runner model will be validated by the hello probe.", + }); + } catch (error) { + checks.push({ code: "opencode_model_invalid", level: "error", + message: error instanceof Error ? error.message : "Configured model is invalid." }); + } + } else if (targetIsRemote && configuredModel) { checks.push({ code: "opencode_model_validation_skipped_remote", level: "info", @@ -305,7 +330,7 @@ export async function testEnvironment( const modelUnavailable = checks.some((check) => check.code === "opencode_hello_probe_model_unavailable"); if (!configuredModel && !modelUnavailable) { // No model configured – skip model requirement if no model-related checks exist - } else if (!targetIsRemote && configuredModel && canRunProbe) { + } else if (!nativeModelProbe && !targetIsRemote && configuredModel && canRunProbe) { try { await ensureOpenCodeModelConfiguredAndAvailable({ model: configuredModel, @@ -430,6 +455,9 @@ export async function testEnvironment( } } } finally { + if (nativeProbeWorkspaceLocalDir) { + await fs.rm(nativeProbeWorkspaceLocalDir, { recursive: true, force: true }).catch(() => {}); + } await restoreWorkspace?.(); if (!restoreWorkspace && preparedRuntimeWorkspaceLocalDir) { // Reached when `prepareAdapterExecutionTargetRuntime` threw before diff --git a/packages/paperclip-runner/README.md b/packages/paperclip-runner/README.md index f87a1d07652..a7f0647a39f 100644 --- a/packages/paperclip-runner/README.md +++ b/packages/paperclip-runner/README.md @@ -238,6 +238,37 @@ provider-native operations. Full auto is resolved consistently for agent creation, adapter conversion, direct driver launches, and fresh/resumed turns. Explicitly stored restrictive modes still apply. +Native OpenCode forwards incremental text and reasoning parts as they arrive; +the final full snapshot selects the final response without repeating streamed +text. Delta parts must match an observed message and part identity. Identical +chunks without event IDs remain distinct tokens. Each isolated launch disables +automatic updates, remote model-catalog refreshes, and default plugins; the +runner supplies the pinned executable, selected model, and assigned MCP tools. +Provider inference and reasoning still contribute to time to first text. These +startup settings do not disable model reasoning. + +On macOS, the verified OpenCode executable keeps its private launch pathname +until the provider process exits. Removing that pathname immediately after +spawn can kill or stall the signed binary before its health check succeeds. +Ownership, permissions, and file-identity checks remain enforced; cleanup also +runs on failed launches, and retries create a fresh launch snapshot. + +Call `session.startTurn({ message, reasoningMode: "disabled" })` to disable +reasoning for one OpenCode/OpenRouter turn. `CapabilityLiveSession.sendMessage` +accepts the same option in its second argument. The choice travels in the durable +`turn.start` command and selects an OpenCode model variant for that prompt only. +Omitting it (or passing `"default"`) on the next turn restores provider defaults, +including on a warm or recovered session. It is not an environment variable or +agent-wide setting. No automatic decision policy selects it. The model must +support disabling reasoning; other runner providers reject an explicit selection +before starting work. Native sessions expose support through the +`perTurnReasoning` capability. + +OpenCode's proxy sends agent instructions once per turn through the system prompt. The +task envelope retains task-specific constraints and the completion contract, +without repeating the system instructions. This applies to all users of the +proxy; already-recorded conversation history is not rewritten. + The runner's authenticated bridge and controller still enforce company access, action claims, task modes, and governed approvals. Provider permission defaults do not change workspace isolation or grant credentials or connection access. diff --git a/packages/paperclip-runner/runner/crates/runner-core/src/codex_provider.rs b/packages/paperclip-runner/runner/crates/runner-core/src/codex_provider.rs index 29494a4e21a..61432b9dbe0 100644 --- a/packages/paperclip-runner/runner/crates/runner-core/src/codex_provider.rs +++ b/packages/paperclip-runner/runner/crates/runner-core/src/codex_provider.rs @@ -386,6 +386,13 @@ pub struct CodexSkillInput { pub path: String, } +/// Optional inputs for one turn; omitted fields preserve provider defaults. +#[derive(Clone, Copy, Debug, Default)] +pub struct CodexTurnOptions<'a> { + pub skills: &'a [CodexSkillInput], + pub reasoning_mode: Option<&'a str>, +} + impl CodexSkillInput { pub fn validate(&self) -> Result<(), LocalRunnerError> { if self.input_type != "skill" @@ -1670,15 +1677,26 @@ impl CodexProvider { } pub fn start_turn(&mut self, message: &str, cwd: &str) -> Result { - self.start_turn_with_skills(message, cwd, &[]) + self.start_turn_with_options(message, cwd, CodexTurnOptions::default()) } - pub fn start_turn_with_skills( + pub fn start_turn_with_options( &mut self, message: &str, cwd: &str, - skills: &[CodexSkillInput], + options: CodexTurnOptions<'_>, ) -> Result { + let CodexTurnOptions { + skills, + reasoning_mode, + } = options; + if reasoning_mode.is_some_and(|mode| { + self.config.provider != "opencode" || !matches!(mode, "default" | "disabled") + }) { + return Err(LocalRunnerError::invalid( + "turn.reasoningMode requires OpenCode and must be default or disabled", + )); + } if skills.len() > 64 || (self.config.provider != "codex" && !skills.is_empty()) { return Err(LocalRunnerError::invalid( "explicit skills require Codex and at most 64 selections", @@ -1727,6 +1745,9 @@ impl CodexProvider { let turn_params_object = turn_params .as_object_mut() .expect("Codex turn parameters are an object"); + if let Some(mode) = reasoning_mode { + turn_params_object.insert("reasoningMode".to_owned(), json!(mode)); + } if self.permission_profile == "paperclip-runner-external-sandbox" { turn_params_object.insert( "sandboxPolicy".to_owned(), diff --git a/packages/paperclip-runner/runner/crates/runner-core/src/provider_backend.rs b/packages/paperclip-runner/runner/crates/runner-core/src/provider_backend.rs index 23268296bd6..329c3dc7bf4 100644 --- a/packages/paperclip-runner/runner/crates/runner-core/src/provider_backend.rs +++ b/packages/paperclip-runner/runner/crates/runner-core/src/provider_backend.rs @@ -13,7 +13,7 @@ use serde_json::{json, Value}; use sha2::{Digest, Sha256}; use crate::codex_provider::{ - CodexProvider, CodexProviderConfig, CodexProviderEvent, CodexSkillInput, + CodexProvider, CodexProviderConfig, CodexProviderEvent, CodexSkillInput, CodexTurnOptions, ProviderStartupObservation, ProviderStartupStage, RejectedAcceptedTurn, MAX_SETTLED_PROVIDER_TURN_IDS, }; @@ -2903,7 +2903,24 @@ impl CodexCommandExecutor { .validate() .map_err(|error| DurableRunnerError::invalid(error.to_string()))?; } + let reasoning_mode = match payload.get("reasoningMode") { + None => None, + Some(Value::String(mode)) if matches!(mode.as_str(), "default" | "disabled") => { + Some(mode.as_str()) + } + Some(_) => return Err(DurableRunnerError::invalid("invalid turn.reasoningMode")), + }; self.restore_provider_if_needed()?; + if reasoning_mode.is_some() + && self + .state + .as_ref() + .is_none_or(|state| state.config.provider != "opencode") + { + return Err(DurableRunnerError::invalid( + "turn.reasoningMode requires OpenCode", + )); + } if self .state .as_ref() @@ -2986,7 +3003,14 @@ impl CodexCommandExecutor { rejected_accepted_turn, ) = { let provider = self.ensure_provider()?; - let result = provider.start_turn_with_skills(text, &cwd, &skills); + let result = provider.start_turn_with_options( + text, + &cwd, + CodexTurnOptions { + skills: &skills, + reasoning_mode, + }, + ); ( result, provider.completed_turn_authority().is_some(), @@ -6068,6 +6092,27 @@ mod tests { executor.state = Some(state); executor.restore_checked = true; + for mode in [ + json!(null), + json!(false), + json!("bad"), + json!("disabled"), + json!("default"), + ] { + let error = executor + .start_turn(&json!({"text": "must not reach the provider", "reasoningMode": mode})) + .unwrap_err(); + assert!(error.to_string().contains("reasoningMode")); + assert!(executor.provider.is_none()); + assert!( + !executor + .state + .as_ref() + .unwrap() + .ambiguous_turn_start_pending + ); + } + let error = executor .start_turn(&json!({"text": "must not reach the provider"})) .expect_err("the exact regular backlog limit must reject admission"); diff --git a/packages/paperclip-runner/src/backends/harness-driver-backend.test.ts b/packages/paperclip-runner/src/backends/harness-driver-backend.test.ts index e1542606a24..a5cc423e60e 100644 --- a/packages/paperclip-runner/src/backends/harness-driver-backend.test.ts +++ b/packages/paperclip-runner/src/backends/harness-driver-backend.test.ts @@ -97,6 +97,34 @@ const driver: HarnessDriver = { }; describe("HarnessDriverBackend", () => { + it.each(["default", "disabled"] as const)("rejects unsupported reasoning %s before starting provider work", async reasoningMode => { + const startTurn = vi.fn(async () => ({ turnId: "turn-1" })); + const session = Object.assign(new FakeHarnessSession(), { startTurn }); + const backend = new HarnessDriverBackend({ ...driver, openSession: async () => session }); + const opened = await backend.openSession({ identity: { + runId: "run-1", sessionId: "session-1", companyId: "company-1", issueId: "issue-1", agentId: "agent-1", + } }); + expect((await opened.capabilities()).perTurnReasoning).toBe(false); + const message = { role: "user" as const, text: "Hi" }; + await expect(opened.startTurn({ message, reasoningMode })).rejects.toThrow("Per-turn reasoning is not supported"); + expect(startTurn).not.toHaveBeenCalled(); + await expect(opened.startTurn({ message })).resolves.toEqual({ turnId: "turn-1" }); + expect(startTurn).toHaveBeenCalledExactlyOnceWith({ message }); + }); + + it("forwards reasoning when the provider explicitly supports it", async () => { + const startTurn = vi.fn(async () => ({ turnId: "turn-1" })); + const session = Object.assign(new FakeHarnessSession(), { startTurn, supportsTurnReasoning: () => true }); + const backend = new HarnessDriverBackend({ ...driver, openSession: async () => session }); + const opened = await backend.openSession({ identity: { + runId: "run-1", sessionId: "session-1", companyId: "company-1", issueId: "issue-1", agentId: "agent-1", + } }); + expect((await opened.capabilities()).perTurnReasoning).toBe(true); + const input = { message: { role: "user" as const, text: "Hi" }, reasoningMode: "disabled" as const }; + await opened.startTurn(input); + expect(startTurn).toHaveBeenCalledExactlyOnceWith(input); + }); + it("refreshes negotiated controls after a warm handshake and separates follow-ups", async () => { let controls = { steering: false, queuedFollowUp: false }; const steer = vi.fn(async () => {}); diff --git a/packages/paperclip-runner/src/backends/harness-driver-backend.ts b/packages/paperclip-runner/src/backends/harness-driver-backend.ts index 6e3897e4e57..fcdb43e16b7 100644 --- a/packages/paperclip-runner/src/backends/harness-driver-backend.ts +++ b/packages/paperclip-runner/src/backends/harness-driver-backend.ts @@ -454,6 +454,7 @@ class HarnessNativeSession implements NativeSession { return { resume: true, typedEvents: true, + perTurnReasoning: this.#session.supportsTurnReasoning?.() === true, steering: (negotiated?.steering ?? this.#steeringSupported) && this.#session.steer !== undefined, queuedFollowUp: negotiated?.queuedFollowUp === true && this.#session.steer !== undefined, interruption: this.#session.interrupt !== undefined, @@ -710,6 +711,9 @@ class HarnessNativeSession implements NativeSession { // The provider has no active turn to interrupt yet. Do not launch work // whose events cancellation would suppress and leave the owner waiting. if (this.#explicitlyCancelled) throw new Error("native_session_cancelled"); + if (input.reasoningMode !== undefined && this.#session.supportsTurnReasoning?.() !== true) { + throw new Error("Per-turn reasoning is not supported by this provider"); + } this.#terminal = null; try { const started = await this.#session.startTurn(input); diff --git a/packages/paperclip-runner/src/backends/native-instruction-measurement.test.ts b/packages/paperclip-runner/src/backends/native-instruction-measurement.test.ts index 98507b24338..2c5592af47e 100644 --- a/packages/paperclip-runner/src/backends/native-instruction-measurement.test.ts +++ b/packages/paperclip-runner/src/backends/native-instruction-measurement.test.ts @@ -190,11 +190,10 @@ describe("direct OpenCode HTTP instruction boundary", () => { const request = requests.at(-1)!; expect(request.providerID).toBe("openrouter"); expect(request.modelID).toBe("deepseek/deepseek-v4-flash-0731"); const text = request.parts[0].text as string; + expect(request.system).toBe(nativeSystemInstructions(value)); if (schema === "v4" && phase === "start") { - expect(request.system).toBe(nativeSystemInstructions(input)); expect(JSON.parse(text).task.constraints).toEqual(nativeTaskConstraints(input)); } else { - expect(request.system).toBeUndefined(); expect(JSON.parse(text)).toEqual(envelope); } directOpenCodeReceipts.push({ provider: "opencode", schema, phase, request: measure(request), diff --git a/packages/paperclip-runner/src/cli/opencode-app-server-proxy.test.ts b/packages/paperclip-runner/src/cli/opencode-app-server-proxy.test.ts index 0c672b356af..65250a0f8d5 100644 --- a/packages/paperclip-runner/src/cli/opencode-app-server-proxy.test.ts +++ b/packages/paperclip-runner/src/cli/opencode-app-server-proxy.test.ts @@ -9,7 +9,7 @@ import { rmSync, writeFileSync, } from "node:fs"; -import { spawnSync } from "node:child_process"; +import { spawn, spawnSync } from "node:child_process"; import { tmpdir } from "node:os"; import { join } from "node:path"; import { @@ -28,7 +28,7 @@ describe("OpenCode runnerd proxy executable", () => { ])).toEqual({ command: "/proc/self/fd/7", commandFd: 7 }); }); - it("executes the inherited artifact through the nested child fd mapping", () => { + it("executes the inherited artifact through the nested child fd mapping", async () => { if (process.platform === "win32") return; if (process.platform === "darwin") { const root = mkdtempSync(join(tmpdir(), "paperclip-opencode-nested-")); @@ -45,22 +45,27 @@ describe("OpenCode runnerd proxy executable", () => { TRUSTED_OPENCODE_EXECUTABLE_ARG, command, ]); - binding.commandLifecycle?.beforeSpawn(); - const child = spawnSync(binding.command, [], { encoding: "utf8" }); - expect(child.error).toBeUndefined(); - expect(child.status).toBe(0); - expect(child.stdout).toBe("verified-nested-spawn"); - binding.commandLifecycle?.afterSpawn(); - expect(() => lstatSync(command)).toThrow(); - expect(() => lstatSync(directory)).toThrow(); - binding.commandLifecycle?.beforeSpawn(); - const retriedChild = spawnSync(binding.command, [], { encoding: "utf8" }); - expect(retriedChild.error).toBeUndefined(); - expect(retriedChild.status).toBe(0); - expect(retriedChild.stdout).toBe("verified-nested-spawn"); - binding.commandLifecycle?.afterSpawn(); - expect(() => lstatSync(command)).toThrow(); - expect(() => lstatSync(directory)).toThrow(); + for (let attempt = 0; attempt < 2; attempt += 1) { + binding.commandLifecycle!.beforeSpawn(); + const child = spawn(binding.command, [], { stdio: ["ignore", "pipe", "pipe"] }); + let stdout = ""; + child.stdout!.on("data", chunk => { stdout += chunk; }); + const exited = new Promise((resolve, reject) => { + child.once("error", reject); + child.once("close", resolve); + }); + binding.commandLifecycle!.afterSpawn(); + // Exercise the production async boundary, not spawnSync followed by + // cleanup after the executable has already finished. + expect(lstatSync(command).isFile()).toBe(true); + expect(() => binding.commandLifecycle!.beforeSpawn()).toThrow("still in use"); + expect(await exited).toBe(0); + expect(stdout).toBe("verified-nested-spawn"); + binding.commandLifecycle!.afterExit(); + binding.commandLifecycle!.afterExit(); + expect(() => lstatSync(command)).toThrow(); + expect(() => lstatSync(directory)).toThrow(); + } } finally { rmSync(root, { recursive: true, force: true }); } @@ -131,6 +136,7 @@ describe("OpenCode runnerd proxy task envelope", () => { criterionIds: ["objective", "verification"], }, })).toMatchObject({ + constraints: ["Use Paperclip MCP tools for semantic operations."], completionContract: { revision: "17", criteria: [ diff --git a/packages/paperclip-runner/src/cli/opencode-app-server-proxy.ts b/packages/paperclip-runner/src/cli/opencode-app-server-proxy.ts index 11cc112dea7..0e60ac1f245 100644 --- a/packages/paperclip-runner/src/cli/opencode-app-server-proxy.ts +++ b/packages/paperclip-runner/src/cli/opencode-app-server-proxy.ts @@ -1,4 +1,5 @@ #!/usr/bin/env node +import { parseOpenCodeReasoningMode } from "../drivers/opencode/reasoning-mode.js"; import { createInterface } from "node:readline"; import { readFileSync } from "node:fs"; import { resolve } from "node:path"; @@ -365,6 +366,7 @@ async function handle(message: RpcMessage): Promise { .join("\n"); const turn = await session.startTurn({ message: { role: "user", text: messageText }, + ...(params.reasoningMode === undefined ? {} : { reasoningMode: parseOpenCodeReasoningMode(params.reasoningMode) }), }); activeTurnId = turn.turnId; // OpenCode normally publishes session/turn startup over SSE, but a fast diff --git a/packages/paperclip-runner/src/cli/opencode-proxy-command.ts b/packages/paperclip-runner/src/cli/opencode-proxy-command.ts index 2da3349bf34..3d73ce8d22b 100644 --- a/packages/paperclip-runner/src/cli/opencode-proxy-command.ts +++ b/packages/paperclip-runner/src/cli/opencode-proxy-command.ts @@ -39,6 +39,7 @@ export function trustedOpenCodeLaunchBinding( commandLifecycle?: { beforeSpawn(): void; afterSpawn(): void; + afterExit(): void; }; } { const command = args.length === 2 && args[0] === TRUSTED_OPENCODE_EXECUTABLE_ARG @@ -119,8 +120,7 @@ export function trustedOpenCodeLaunchBinding( let materializedIdentity: { dev: number; ino: number } | undefined; const materializeForSpawn = () => { if (materialized) { - validateMacSnapshot(materializedIdentity); - return; + throw new Error("Verified OpenCode executable is still in use"); } let destinationFd: number | undefined; try { @@ -184,10 +184,15 @@ export function trustedOpenCodeLaunchBinding( // attackers are outside the documented local-host trust boundary in // docs/durable-recovery.md. Keep the verified source on an unlinked // descriptor, rematerialize only at the syscall boundary (including - // retries), then remove the executable pathname immediately. + // retries). Keep its private pathname until the child exits: deleting + // a signed executable during macOS startup can kill or stall the child. beforeSpawn: materializeForSpawn, afterSpawn() { validateMacSnapshot(materializedIdentity); + }, + afterExit() { + if (!materialized) return; + validateMacSnapshot(materializedIdentity); unlinkSync(command); rmdirSync(dirname(command)); materialized = false; diff --git a/packages/paperclip-runner/src/cli/opencode-proxy-task-envelope.ts b/packages/paperclip-runner/src/cli/opencode-proxy-task-envelope.ts index dd30c666d90..535b64d6d30 100644 --- a/packages/paperclip-runner/src/cli/opencode-proxy-task-envelope.ts +++ b/packages/paperclip-runner/src/cli/opencode-proxy-task-envelope.ts @@ -30,7 +30,8 @@ export function openCodeProxyTaskEnvelope(params: Record) { } : {}), constraints: [ - text(params.baseInstructions, "Complete only the supplied task."), + // baseInstructions already travels in the system prompt. Keep task + // constraints here without duplicating the agent instruction bundle. "Use Paperclip MCP tools for semantic operations.", ], }); diff --git a/packages/paperclip-runner/src/contracts/harness-driver.ts b/packages/paperclip-runner/src/contracts/harness-driver.ts index 410496afce5..a520a146431 100644 --- a/packages/paperclip-runner/src/contracts/harness-driver.ts +++ b/packages/paperclip-runner/src/contracts/harness-driver.ts @@ -498,6 +498,7 @@ export interface HarnessSessionRecoveryResult { } export interface HarnessSession { + supportsTurnReasoning?(): boolean; turnControlCapabilities?(): NativeTurnControlCapabilities | null; ids(): { driverSessionId: string; @@ -511,6 +512,8 @@ export interface HarnessSession { /** Set by orchestration only after successful provider-session recovery. */ continuation?: true; requestedCollaborationMode?: "default" | "plan"; + /** OpenCode/OpenRouter only. Applies to this turn, never subsequent turns. */ + reasoningMode?: "default" | "disabled"; }): Promise<{ turnId: string; effectiveCollaborationMode?: "default" | "plan"; diff --git a/packages/paperclip-runner/src/contracts/native-session-backend.ts b/packages/paperclip-runner/src/contracts/native-session-backend.ts index f473b56a03e..251d10748e7 100644 --- a/packages/paperclip-runner/src/contracts/native-session-backend.ts +++ b/packages/paperclip-runner/src/contracts/native-session-backend.ts @@ -193,6 +193,8 @@ export interface NativeSession { /** Set by orchestration only after successful provider-session recovery. */ continuation?: true; requestedCollaborationMode?: "default" | "plan"; + /** OpenCode/OpenRouter only. Applies to this turn, never subsequent turns. */ + reasoningMode?: "default" | "disabled"; }): Promise<{ turnId: string; effectiveCollaborationMode?: "default" | "plan"; diff --git a/packages/paperclip-runner/src/contracts/types.ts b/packages/paperclip-runner/src/contracts/types.ts index 81bc3d24723..8a5dd2a49bd 100644 --- a/packages/paperclip-runner/src/contracts/types.ts +++ b/packages/paperclip-runner/src/contracts/types.ts @@ -41,6 +41,8 @@ export interface NativeSessionCapabilities { runtimeRequestHandoff?: boolean; goals?: boolean; threadLineage?: boolean; + /** Can select reasoning for one turn without changing subsequent turns. */ + perTurnReasoning?: boolean; collaborationModes?: Array<"default" | "plan">; unsupported?: string[]; } diff --git a/packages/paperclip-runner/src/drivers/codex/app-server-transport.ts b/packages/paperclip-runner/src/drivers/codex/app-server-transport.ts index 3e08de1f22f..a65138a435b 100644 --- a/packages/paperclip-runner/src/drivers/codex/app-server-transport.ts +++ b/packages/paperclip-runner/src/drivers/codex/app-server-transport.ts @@ -35,6 +35,7 @@ export type CodexServerRequestHandler = ( ) => Promise>; export interface CodexAppServerTransport { + supportsTurnReasoning?(): boolean; /** Runner-owned live capability projection; absent on native Codex transports. */ turnControlCapabilities?(): NativeTurnControlCapabilities | null; request( diff --git a/packages/paperclip-runner/src/drivers/codex/codex-app-server-driver.context-control.test.ts b/packages/paperclip-runner/src/drivers/codex/codex-app-server-driver.context-control.test.ts index cd9ee34129d..021c0ab99f7 100644 --- a/packages/paperclip-runner/src/drivers/codex/codex-app-server-driver.context-control.test.ts +++ b/packages/paperclip-runner/src/drivers/codex/codex-app-server-driver.context-control.test.ts @@ -45,6 +45,19 @@ import { import { rehydrateRunnerdItemNotification } from "../../live/runnerd-codex-transport.js"; describe("Codex app-server Codex driver", () => { + it.each(["default", "disabled"] as const)("rejects unsupported reasoning %s without sending a turn RPC", async reasoningMode => { + const transport = new FakeCodexTransport(); + const session = await makeDriver([transport]).openSession({ + runId: "run-reasoning", normalizedSessionId: "normalized-reasoning", workingDirectory: WORKSPACE, + }); + const message = { role: "user" as const, text: "Hi" }; + await expect(session.startTurn({ message, reasoningMode })).rejects.toThrow("Per-turn reasoning is not supported"); + expect(transport.calls.filter(call => call.method === "turn/start")).toHaveLength(0); + await session.startTurn({ message }); + expect(transport.calls.filter(call => call.method === "turn/start")).toHaveLength(1); + await session.close({ reason: "fixture complete" }); + }); + it("retains one correlation-bound steering acknowledgement after the rehydrated runnerd echo", async () => { const transport = new FakeCodexTransport("thread-root"); const session = await makeDriver([transport]).openSession({ diff --git a/packages/paperclip-runner/src/drivers/codex/codex-harness-session.ts b/packages/paperclip-runner/src/drivers/codex/codex-harness-session.ts index 4efc9eec23d..983d9322eee 100644 --- a/packages/paperclip-runner/src/drivers/codex/codex-harness-session.ts +++ b/packages/paperclip-runner/src/drivers/codex/codex-harness-session.ts @@ -82,6 +82,10 @@ export class CodexHarnessSession } } + supportsTurnReasoning(): boolean { + return this.transport.supportsTurnReasoning?.() === true; + } + turnControlCapabilities() { if (this.driverKind === "acpx_runtime") { return this.transport.turnControlCapabilities?.() ?? { steering: false, queuedFollowUp: false }; @@ -157,11 +161,16 @@ export class CodexHarnessSession /** Set by orchestration only after successful provider-session recovery. */ continuation?: true; requestedCollaborationMode?: "default" | "plan"; + /** OpenCode/OpenRouter only. Applies to this turn, never subsequent turns. */ + reasoningMode?: "default" | "disabled"; }): Promise<{ turnId: string; effectiveCollaborationMode: "default" | "plan"; }> { this.assertProtocolIntegrity(); + if (input.reasoningMode !== undefined && !this.supportsTurnReasoning()) { + throw new Error("Per-turn reasoning is not supported by this provider"); + } if (this.protocolFailed && this.protocolFailureCode) { throw new NativeProviderTerminalFailure(this.protocolFailureCode, false, this.protocolFailureMessage ?? undefined); } @@ -231,6 +240,7 @@ export class CodexHarnessSession try { response = await this.transport.request("turn/start", { threadId: this.opened.threadId, + ...(input.reasoningMode === undefined ? {} : { reasoningMode: input.reasoningMode }), ...(this.reasoningEffort ? { effort: this.reasoningEffort } : {}), cwd: this.opened.context.workingDirectory, permissions: diff --git a/packages/paperclip-runner/src/drivers/opencode/opencode-server-driver.test.ts b/packages/paperclip-runner/src/drivers/opencode/opencode-server-driver.test.ts index 59d740d9940..75032bd82c7 100644 --- a/packages/paperclip-runner/src/drivers/opencode/opencode-server-driver.test.ts +++ b/packages/paperclip-runner/src/drivers/opencode/opencode-server-driver.test.ts @@ -906,6 +906,8 @@ describe("OpenCodeServerDriver", () => { expect(environment.keys).not.toContain("UNRELATED_SECRET"); expect(environment.keys).not.toContain("PAPERCLIP_PROVIDER_TRACE_PATH"); expect(environment.projectConfigDisabled).toBe("true"); + expect(environment.modelsFetchDisabled).toBe("true"); + expect(environment.defaultPluginsDisabled).toBe("true"); expect(mcpEvidence.tools).toEqual( expect.arrayContaining(["paperclip_finish", "paperclip_block"]), ); @@ -920,6 +922,46 @@ describe("OpenCodeServerDriver", () => { expect(diagnostics.join("\n")).toContain("[REDACTED]"); }); + it.each(["role-first", "role-last", "role-last-long"])("preserves incremental text and completion with identity checks (%s)", async (roleOrder) => { + await chmod(fixture, 0o755); + const root = await mkdtemp(join(tmpdir(), "paperclip-opencode-stream-")); + const workspace = await mkdtemp(join(tmpdir(), "paperclip-opencode-stream-workspace-")); + roots.push(root, workspace); + const driver = new OpenCodeServerDriver({ + model: "openrouter/deepseek/deepseek-v4-flash-0731", + runtimeDirectory: root, + command: fixture, + environment: { PATH: process.env.PATH }, + }); + const session = await driver.openSession({ runId: "stream", normalizedSessionId: "stream", workingDirectory: workspace }); + try { + await session.startTurn({ message: { role: "user", text: `incremental-stream ${roleOrder}` } }); + const deltas: string[] = []; + const reasoning: string[] = []; + const events: PrpEvent[] = []; + for await (const event of session.events()) { + events.push(event); + if (event.eventType === "item.delta" && event.payload.kind === "reasoning") reasoning.push(String(event.payload.text)); + if (event.eventType === "item.delta" && event.payload.kind === "agentMessage") { + deltas.push(String(event.payload.text)); + // The fixture cannot finish its turn until a chunk reaches us. + if (deltas.length === 1) { + expect(await readFile(join(root, "stream", "data", "stream-released"), "utf8").catch(() => null)).toBeNull(); + } + await writeFile(join(root, "stream", "data", "release-stream"), "ready"); + } + if (TURN_TERMINAL_EVENT_TYPES.has(event.eventType)) break; + } + const expectedText = `Hi ${"👋".repeat(roleOrder === "role-last-long" ? 120 : 2)}`; + expect(deltas).toEqual(roleOrder === "role-first" ? ["Hi ", "👋", "👋"] : [expectedText]); + expect(reasoning).toEqual(["Thinking"]); + expect(events.filter(event => event.eventType === "item.completed" && event.payload.kind === "agentMessage") + .map(event => event.payload.text)).toEqual([expectedText]); + } finally { + await session.close({ reason: "stream test complete" }); + } + }); + it("maps OpenCode's normal abort error to cancellation without a false provider failure notice", async () => { await chmod(fixture, 0o755); const root = await mkdtemp(join(tmpdir(), "paperclip-opencode-abort-")); @@ -1497,7 +1539,7 @@ describe("OpenCodeServerDriver", () => { await session.close({ reason: "test" }); }); - it("sends only the authoritative wake envelope when resuming an existing provider session", async () => { + it("resends system instructions without repeating the task envelope on recovery", async () => { await chmod(fixture, 0o755); const root = await mkdtemp( join(tmpdir(), "paperclip-opencode-resume-context-"), @@ -1553,7 +1595,7 @@ describe("OpenCodeServerDriver", () => { ], }); expect(submittedPrompt).not.toHaveProperty("tools"); - expect(submittedPrompt).not.toHaveProperty("system"); + expect(submittedPrompt).toHaveProperty("system", "large original system context"); await recovered!.session!.close({ reason: "recovery-test" }); }); @@ -1942,9 +1984,11 @@ describe("OpenCodeServerDriver", () => { "*": permissionMode, external_directory: { "*": "deny", [`${workspace}/**`]: "allow" }, }); + expect(config.autoupdate).toBe(false); expect(config.provider.openrouter.models).toHaveProperty( "deepseek/deepseek-v4-flash-0731", ); + expect(config.provider.openrouter.models["deepseek/deepseek-v4-flash-0731"].options).toBeUndefined(); expect(config.permission.paperclip_finish).toBeUndefined(); expect(config.permission["paperclip_*"]).toBe("allow"); await session.close({ reason: "permission mode test complete" }); @@ -2440,6 +2484,9 @@ describe("OpenCodeServerDriver", () => { afterSpawn: () => { commandLifecycle.push("after"); }, + afterExit: () => { + commandLifecycle.push("exit"); + }, }, fetch: async (input, init) => { if ( @@ -2465,9 +2512,31 @@ describe("OpenCodeServerDriver", () => { expect(spawns.length).toBeGreaterThanOrEqual(2); expect(spawns.length).toBeLessThanOrEqual(3); expect(commandLifecycle).toEqual( - Array.from({ length: spawns.length }, () => ["before", "after"]).flat(), + [...Array.from({ length: spawns.length - 1 }, () => ["before", "after", "exit"]).flat(), "before", "after"], ); await session.close({ reason: "test" }); + expect(commandLifecycle).toEqual( + Array.from({ length: spawns.length }, () => ["before", "after", "exit"]).flat(), + ); + }); + + it("releases the executable binding when spawning fails before health checks", async () => { + const root = await mkdtemp(join(tmpdir(), "paperclip-opencode-spawn-error-")); + roots.push(root); + const lifecycle: string[] = []; + const driver = new OpenCodeServerDriver({ + model: "openrouter/deepseek/deepseek-v4-flash-0731", + runtimeDirectory: root, + command: join(root, "missing-executable"), + commandLifecycle: { + beforeSpawn: () => { lifecycle.push("before"); }, + afterSpawn: () => { lifecycle.push("spawned"); }, + afterExit: () => { lifecycle.push("released"); }, + }, + }); + await expect(driver.openSession({ runId: "spawn-error", normalizedSessionId: "spawn-error", + workingDirectory: root })).rejects.toThrow("ENOENT"); + expect(lifecycle).toEqual(["before", "released"]); }); it("reports startup exit details with stderr captured after spawn and redacted", async () => { diff --git a/packages/paperclip-runner/src/drivers/opencode/opencode-server-driver.ts b/packages/paperclip-runner/src/drivers/opencode/opencode-server-driver.ts index 81abbfd3f06..4ce6a5b9dd9 100644 --- a/packages/paperclip-runner/src/drivers/opencode/opencode-server-driver.ts +++ b/packages/paperclip-runner/src/drivers/opencode/opencode-server-driver.ts @@ -19,6 +19,8 @@ import { getCACertificates } from "node:tls"; import { pipeline } from "node:stream/promises"; import { dirname, isAbsolute, join, relative, resolve, sep } from "node:path"; +import { parseOpenCodeReasoningMode, type OpenCodeReasoningMode } from "./reasoning-mode.js"; + import { CODEX_SKILLLESS_BASE_INSTRUCTIONS, createCodexTaskEnvelope, @@ -176,6 +178,7 @@ export interface OpenCodeServerDriverOptions { commandLifecycle?: { beforeSpawn(): void; afterSpawn(): void; + afterExit?(): void; }; runtimeDirectory: string; systemInstructions?: string; @@ -479,6 +482,7 @@ class OpenCodeHarnessSession implements HarnessSession { Array> >(); readonly #partText = new Map(); + readonly #streamingParts = new Map>(); readonly #completedTextPartIds = new Set(); readonly #completedReasoningPartIds = new Set(); readonly #completedTextParts: Array<{ @@ -640,12 +644,22 @@ class OpenCodeHarnessSession implements HarnessSession { .finally(() => this.#pumpEvents()); } + supportsTurnReasoning(): boolean { + return this.#model.startsWith("openrouter/"); + } + async startTurn(input: { message: NativeUserMessage; + reasoningMode?: OpenCodeReasoningMode; }): Promise<{ turnId: string }> { if (this.#activeTurnId !== null) throw new Error("OpenCode session already has an active turn"); + const reasoningMode = parseOpenCodeReasoningMode(input.reasoningMode); + if (input.reasoningMode !== undefined && !this.#model.startsWith("openrouter/")) { + throw new Error("Per-turn OpenCode reasoning is supported only for OpenRouter models"); + } const turnId = `turn-${randomBytes(12).toString("hex")}`; + this.#streamingParts.clear(); this.#activeTurnId = turnId; this.#emit("turn.submitted", { envelopeSchema: this.#taskEnvelope.schema, @@ -655,12 +669,9 @@ class OpenCodeHarnessSession implements HarnessSession { this.#emit("turn.started", { status: "inProgress" }, { turnId }); const [providerID, ...modelParts] = this.#model.split("/"); const modelID = modelParts.join("/"); - // A resumed OpenCode provider session already retains the original system - // instructions and task envelope in its conversation. Repeating both on - // every Paperclip continuation can overflow smaller context windows and - // OpenCode then completes with `finish: unknown` and zero tokens. The - // native model envelope still carries the authoritative wake delta, - // interaction responses, completion contract, and current issue context. + // Keep the task envelope only on the initial task-mode wake. OpenCode + // rebuilds system instructions from the latest user message, so system + // instructions must still accompany every prompt (including recovery). const prompt = this.#sendFullContext ? JSON.stringify({ task: this.#taskEnvelope, @@ -682,12 +693,15 @@ class OpenCodeHarnessSession implements HarnessSession { // at this HTTP boundary. providerID, modelID, + // Always select the empty default variant on an ordinary turn so a + // previous disabled turn cannot change this turn's provider defaults. + ...(providerID === "openrouter" + ? { variant: reasoningMode === "disabled" ? "paperclip-no-reasoning" : "paperclip-default" } + : {}), // Keep question enabled in the isolated config. OpenCode 1.18.32 // turns a prompt's deprecated `tools` map into replacement session // permissions, so a sparse override here discards the session policy. - ...(this.#sendFullContext - ? { system: this.#systemInstructions } - : {}), + system: this.#systemInstructions, parts: [{ type: "text", text: prompt }], }), }, @@ -1470,9 +1484,11 @@ class OpenCodeHarnessSession implements HarnessSession { const event = record(value); const properties = record(event.properties); const type = text(event.type); - const eventId = text(event.id, canonicalJson(value)); - if (this.#seenProviderEvents.has(eventId)) return; - this.#seenProviderEvents.add(eventId); + // Id-less deltas are additive: two identical chunks can be legitimate + // adjacent tokens. Content deduplication would silently lose the second. + const eventId = text(event.id) || (type === "message.part.delta" ? null : canonicalJson(value)); + if (eventId !== null && this.#seenProviderEvents.has(eventId)) return; + if (eventId !== null) this.#seenProviderEvents.add(eventId); if (this.#seenProviderEvents.size > 10_000) this.#seenProviderEvents.delete( this.#seenProviderEvents.values().next().value!, @@ -1565,8 +1581,16 @@ class OpenCodeHarnessSession implements HarnessSession { ); return; } - if (type === "message.part.updated") { - const part = record(properties.part); + if (type === "message.part.updated" || type === "message.part.delta") { + let part = record(properties.part); + if (type === "message.part.delta") { + const previous = this.#streamingParts.get(text(properties.partID)); + // A delta has no role or part type. Only stream text for an observed + // part with matching message identity; never infer assistant identity. + if (!previous || properties.field !== "text" || typeof properties.delta !== "string" + || text(previous.messageID, text(previous.messageId)) !== text(properties.messageID)) return; + part = { ...previous, text: text(previous.text) + properties.delta }; + } const messageId = text(part.messageID, text(part.messageId)); if (!messageId) return; // Resolve the turn this message actually belongs to, not whichever @@ -1574,11 +1598,20 @@ class OpenCodeHarnessSession implements HarnessSession { // must stay attributed to the turn that created that message. const owningTurnId = this.#messageTurnIds.get(messageId) ?? turnId; if (!owningTurnId) return; + if (owningTurnId === turnId && text(part.id) && ["text", "reasoning"].includes(text(part.type))) { + this.#streamingParts.set(text(part.id), part); + } const role = this.#messageRoles.get(messageId); if (role === "assistant") this.#emitAssistantPart(part, owningTurnId); else if (role === undefined) { const pending = this.#pendingMessageParts.get(messageId) ?? []; - if (pending.length < 100) pending.push(part); + // Text deltas are cumulative by this point. Keep the newest snapshot + // (including completion metadata), not the first 100 token chunks. + const previousIndex = ["text", "reasoning"].includes(text(part.type)) && text(part.id) + ? pending.findIndex(candidate => candidate.id === part.id && candidate.type === part.type) + : -1; + if (previousIndex >= 0) pending[previousIndex] = part; + else if (pending.length < 100) pending.push(part); this.#pendingMessageParts.set(messageId, pending); } return; @@ -2233,12 +2266,23 @@ async function startRuntime(input: { input.trace?.addSensitiveValues([providerProxy.token]); } let child: ChildProcess | undefined; + let launchPrepared = false; + const releaseExecutable = () => { + if (!launchPrepared) return; + launchPrepared = false; + try { + input.options.commandLifecycle?.afterExit?.(); + } catch (error) { + input.options.onDiagnostic?.(redact(`OpenCode executable cleanup failed: ${String(error)}`, sensitiveValues)); + } + }; try { const config = { $schema: "https://opencode.ai/config.json", model: input.options.model, small_model: input.options.model, share: "disabled", + autoupdate: false, // The configured entry is already composed exactly once into the session // system prompt; siblings remain available through the read-only root. instructions: [], @@ -2257,7 +2301,15 @@ async function startRuntime(input: { }, } : {}), models: { - [providerModelId]: { name: providerModelId }, + [providerModelId]: { + name: providerModelId, + ...(modelProvider === "openrouter" + ? { variants: { + "paperclip-default": {}, + "paperclip-no-reasoning": { reasoning: { enabled: false } }, + } } + : {}), + }, }, }, }, @@ -2307,6 +2359,11 @@ async function startRuntime(input: { XDG_DATA_HOME: dataHome, XDG_CACHE_HOME: cacheHome, OPENCODE_DISABLE_PROJECT_CONFIG: "true", + // The runner pins the executable, model and assigned MCP tools. A + // fresh isolated cache must not fetch a catalog or install unrelated + // default plugins before it can submit the first prompt. + OPENCODE_DISABLE_MODELS_FETCH: "true", + OPENCODE_DISABLE_DEFAULT_PLUGINS: "true", OPENCODE_SERVER_USERNAME: username, OPENCODE_SERVER_PASSWORD: password, ...(providerProxy ? { NO_PROXY: [input.options.environment?.no_proxy ?? input.options.environment?.NO_PROXY, "127.0.0.1", "localhost"].filter(Boolean).join(",") } : {}), @@ -2319,6 +2376,7 @@ async function startRuntime(input: { stdio[input.options.commandFd] = input.options.commandFd; } input.options.commandLifecycle?.beforeSpawn(); + launchPrepared = true; child = spawn( input.options.command ?? resolvePinnedOpenCodeCommand(), ["serve", "--hostname", "127.0.0.1", "--port", String(port)], @@ -2329,6 +2387,7 @@ async function startRuntime(input: { detached: globalThis.process.platform !== "win32" && isolateProcessGroup, }, ); + child.once("exit", releaseExecutable); if (child.pid !== undefined) { try { input.options.commandLifecycle?.afterSpawn(); @@ -2425,6 +2484,7 @@ async function startRuntime(input: { } catch { providerChild.kill("SIGKILL"); } + await waitForExit(providerChild, 2_000); } await rm(join(configHome, "opencode", "opencode.json"), { force: true, @@ -2441,6 +2501,8 @@ async function startRuntime(input: { await providerProxy?.close(); await bridge.close().catch(() => {}); child?.kill("SIGKILL"); + if (child?.pid) await waitForExit(child, 2_000); + if (!child?.pid || child.exitCode !== null || child.signalCode !== null) releaseExecutable(); await rm(join(configHome, "opencode", "opencode.json"), { force: true, }).catch(() => undefined); diff --git a/packages/paperclip-runner/src/drivers/opencode/prepared-context-transport.test.ts b/packages/paperclip-runner/src/drivers/opencode/prepared-context-transport.test.ts index dd6f34e8677..c4710a971e9 100644 --- a/packages/paperclip-runner/src/drivers/opencode/prepared-context-transport.test.ts +++ b/packages/paperclip-runner/src/drivers/opencode/prepared-context-transport.test.ts @@ -30,6 +30,7 @@ describe("prepared OpenCode context transport", () => { runtimeDirectory: root, command: fixture, conversationMode: "prepared", + systemInstructions: "Prepared agent instructions.", environment: { PATH: process.env.PATH, OPENROUTER_API_KEY: "fixture-key" }, fetch: async (input, init) => { if (String(input).endsWith("/prompt_async")) @@ -45,6 +46,7 @@ describe("prepared OpenCode context transport", () => { }); await session.startTurn({ message: { role: "user", text: "prepared OpenCode initial wake" }, + reasoningMode: "disabled", }); for await (const event of session.events()) { if (event.eventType === "turn.completed") break; @@ -67,6 +69,8 @@ describe("prepared OpenCode context transport", () => { [{ type: "text", text: "completion-only OpenCode continuation" }], ]); expect(JSON.stringify(submitted)).not.toContain('"task"'); + expect(submitted.map(body => body.variant)).toEqual(["paperclip-no-reasoning", "paperclip-default"]); + expect(submitted.map(body => body.system)).toEqual(["Prepared agent instructions.", "Prepared agent instructions."]); await recovered!.session!.close({ reason: "prepared-test-recovery" }); }); }); diff --git a/packages/paperclip-runner/src/drivers/opencode/reasoning-mode.test.ts b/packages/paperclip-runner/src/drivers/opencode/reasoning-mode.test.ts new file mode 100644 index 00000000000..db5e747d46a --- /dev/null +++ b/packages/paperclip-runner/src/drivers/opencode/reasoning-mode.test.ts @@ -0,0 +1,63 @@ +import { chmod, mkdtemp, readFile, rm } from "node:fs/promises"; +import { tmpdir } from "node:os"; +import { join, resolve } from "node:path"; +import { describe, expect, it } from "vitest"; +import { OpenCodeServerDriver } from "./opencode-server-driver.js"; +import { parseOpenCodeReasoningMode } from "./reasoning-mode.js"; +import { openCodeProxyTaskEnvelope } from "../../cli/opencode-proxy-task-envelope.js"; + +describe("per-turn OpenCode reasoning", () => { + it.each([undefined, "default"])("preserves provider defaults for %s", (value) => { + expect(parseOpenCodeReasoningMode(value)).toBe("default"); + }); + it.each(["", "false", null, false, 0])("rejects malformed mode %s", (value) => { + expect(() => parseOpenCodeReasoningMode(value)).toThrow("turn.reasoningMode"); + }); + it("selects reasoning independently on successive turns and sends instructions once", async () => { + const root = await mkdtemp(join(tmpdir(), "paperclip-turn-reasoning-")); + const fixture = resolve("test/fixtures/fake-opencode-server.mjs"); + await chmod(fixture, 0o755); + const submitted: Array<{ variant?: string; system?: string; parts: Array<{ text: string }> }> = []; + const instructions = "Unique shared agent instructions. Keep the requested scope."; + const driver = new OpenCodeServerDriver({ + model: "openrouter/deepseek/deepseek-v4-flash-0731", + runtimeDirectory: root, + command: fixture, + systemInstructions: instructions, + taskEnvelope: openCodeProxyTaskEnvelope({ baseInstructions: instructions, + completionContract: { revision: "17", criterionIds: ["objective"] } }), + environment: { PATH: process.env.PATH, OPENROUTER_API_KEY: "fixture-key" }, + fetch: async (input, init) => { + if (String(input).endsWith("/prompt_async")) submitted.push(JSON.parse(String(init?.body))); + return fetch(input, init); + }, + }); + let session: Awaited> | undefined; + try { + session = await driver.openSession({ runId: "turn-reasoning", normalizedSessionId: "reasoning", workingDirectory: root }); + await expect(session.startTurn({ message: { role: "user", text: "invalid" }, reasoningMode: "bad" as never })).rejects.toThrow("turn.reasoningMode"); + expect(submitted).toHaveLength(0); + for (const reasoningMode of [undefined, "disabled", undefined] as const) { + await session.startTurn({ message: { role: "user", text: "Reply Hi." }, reasoningMode }); + for await (const event of session.events()) if (event.eventType === "turn.completed") break; + } + expect(submitted.map(body => body.variant)).toEqual(["paperclip-default", "paperclip-no-reasoning", "paperclip-default"]); + expect(submitted[0]!.system).toBe(instructions); + for (const body of submitted) { + expect(body.system).toBe(instructions); + expect(JSON.stringify(body).split(instructions)).toHaveLength(2); + } + const envelope = JSON.parse(submitted[0]!.parts[0].text); + expect(envelope.task.completionContract.revision).toBe("17"); + expect(envelope.task.constraints).toEqual(["Use Paperclip MCP tools for semantic operations."]); + const config = JSON.parse(await readFile(join(root, "reasoning/config/opencode/opencode.json"), "utf8")); + expect(config.provider.openrouter.models["deepseek/deepseek-v4-flash-0731"]).toMatchObject({ + variants: { "paperclip-default": {}, "paperclip-no-reasoning": { reasoning: { enabled: false } } }, + }); + expect(config.provider.openrouter.models["deepseek/deepseek-v4-flash-0731"].options).toBeUndefined(); + } finally { + await session?.close(); + await rm(root, { recursive: true, force: true }); + } + }); +}); diff --git a/packages/paperclip-runner/src/drivers/opencode/reasoning-mode.ts b/packages/paperclip-runner/src/drivers/opencode/reasoning-mode.ts new file mode 100644 index 00000000000..003589ead39 --- /dev/null +++ b/packages/paperclip-runner/src/drivers/opencode/reasoning-mode.ts @@ -0,0 +1,8 @@ +export type OpenCodeReasoningMode = "default" | "disabled"; + +/** Opt-in only. Omitting the setting preserves the provider's own defaults. */ +export function parseOpenCodeReasoningMode(value: unknown): OpenCodeReasoningMode { + if (value === undefined || value === "default") return "default"; + if (value === "disabled") return "disabled"; + throw new Error("turn.reasoningMode must be default or disabled"); +} diff --git a/packages/paperclip-runner/src/live/live-session.ts b/packages/paperclip-runner/src/live/live-session.ts index b79755a6685..c36b8036346 100644 --- a/packages/paperclip-runner/src/live/live-session.ts +++ b/packages/paperclip-runner/src/live/live-session.ts @@ -1514,7 +1514,7 @@ export class CapabilityLiveSession { async sendMessage( message: string, /** Launch-only diagnostics may opt out; qualification campaigns must not. */ - options: { allowMissingUsage?: boolean } = {}, + options: { allowMissingUsage?: boolean; reasoningMode?: "default" | "disabled" } = {}, ): Promise { const value = message.trim(); if (value.length === 0) throw new Error("Capability live messages cannot be empty"); @@ -1617,6 +1617,7 @@ export class CapabilityLiveSession { permissions: CODEX_PERMISSION_PROFILE, runtimeWorkspaceRoots: [this.#config.workingDirectory], input: [userInput(value)], + ...(options.reasoningMode === undefined ? {} : { reasoningMode: options.reasoningMode }), }); } catch (error) { if (this.#pendingTurnAdmission !== admission) { diff --git a/packages/paperclip-runner/src/live/runnerd-codex-transport.test.ts b/packages/paperclip-runner/src/live/runnerd-codex-transport.test.ts index a024eb20d92..debd0c6e3fe 100644 --- a/packages/paperclip-runner/src/live/runnerd-codex-transport.test.ts +++ b/packages/paperclip-runner/src/live/runnerd-codex-transport.test.ts @@ -7951,7 +7951,10 @@ it.each([true, false])("preserves prepared OpenCode cleanup errors (primary fail expect((failure as Error).message).not.toContain("fixture-secret"); }); -it("preserves prepared input and completion feedback through runnerd and the real OpenCode proxy boundary", async () => { +it.each([ + { model: "openrouter/deepseek/deepseek-v4-flash-0731", supportsReasoning: true }, + { model: "anthropic/claude-sonnet-4", supportsReasoning: false }, +])("preserves prepared input and completion feedback through runnerd and the real OpenCode proxy boundary ($model)", async ({ model, supportsReasoning }) => { const root = await mkdtemp(join(tmpdir(), "runnerd-prepared-opencode-")); // GitHub-hosted Linux toolcache Node can be group-writable, unlike the AWS // fleet. Qualify an owned copy with strict permissions, never chmod the host @@ -8006,7 +8009,7 @@ it("preserves prepared input and completion feedback through runnerd and the rea const driver = new CodexAppServerDriver({ taskEnvelope: task, conversationMode: "prepared", - model: "openrouter/deepseek/deepseek-v4-flash-0731", + model, transportFactory: () => bundle.transport, workingDirectoryAuthority: "remote_runner", environment: { PAPERCLIP_WORKSPACE_CWD: root }, @@ -8025,7 +8028,15 @@ it("preserves prepared input and completion feedback through runnerd and the rea await withPreparedOpenCodeCleanup({ run: async () => { session = await driver.openSession({ runId: "prepared-opencode", normalizedSessionId: "prepared-opencode", workingDirectory: root }); - await session.startTurn({ message: { role: "user", text: prepared } }); + expect(bundle.transport.supportsTurnReasoning?.()).toBe(supportsReasoning); + const firstTurn = { message: { role: "user" as const, text: prepared } }; + if (!supportsReasoning) { + await expect(session.startTurn({ ...firstTurn, reasoningMode: "disabled" })) + .rejects.toThrow("Per-turn reasoning is not supported"); + await expect(bundle.transport.request("turn/start", { input: [{ type: "text", text: prepared }], reasoningMode: "disabled" })) + .rejects.toThrow("requires an OpenCode/OpenRouter model"); + } + await session.startTurn({ ...firstTurn, ...(supportsReasoning ? { reasoningMode: "disabled" as const } : {}) }); const events: PrpEvent[] = []; for await (const event of session.events()) { events.push(event); @@ -8037,10 +8048,16 @@ it("preserves prepared input and completion feedback through runnerd and the rea expect(sessionRoots).toHaveLength(1); const requests = (await readFile(join(runtime, sessionRoots[0]!.name, "data/fake-prompt-requests.ndjson"), "utf8")).trim().split("\n").map((line) => JSON.parse(line)); expect(requests.map((request) => request.parts)).toEqual([[{ type: "text", text: prepared }]]); + expect(requests[0].variant).toBe(supportsReasoning ? "paperclip-no-reasoning" : undefined); const outcomes = JSON.parse(await readFile(join(runtime, sessionRoots[0]!.name, "data/fake-completion-feedback.json"), "utf8")); expect(outcomes).toHaveLength(2); expect(outcomes[0].result).toMatchObject({ isError: true, content: [{ text: expect.stringContaining("required document link") }] }); expect(outcomes[1].result).toMatchObject({ content: [{ text: expect.stringContaining(feedback) }] }); + await session.startTurn({ message: { role: "user", text: "Reply Hi." } }); + for await (const event of session.events()) if (event.eventType === "turn.completed") break; + const allRequests = (await readFile(join(runtime, sessionRoots[0]!.name, "data/fake-prompt-requests.ndjson"), "utf8")).trim().split("\n").map((line) => JSON.parse(line)); + expect(allRequests.map(request => request.variant)).toEqual(supportsReasoning + ? ["paperclip-no-reasoning", "paperclip-default"] : [undefined, undefined]); }, closeSession: async () => { await session?.close(); }, closeTransport: () => bundle.transport.close(), diff --git a/packages/paperclip-runner/src/live/runnerd-codex-transport.ts b/packages/paperclip-runner/src/live/runnerd-codex-transport.ts index 08a8886f3b9..71103812f2a 100644 --- a/packages/paperclip-runner/src/live/runnerd-codex-transport.ts +++ b/packages/paperclip-runner/src/live/runnerd-codex-transport.ts @@ -3538,6 +3538,20 @@ class DurablePrpCodexTransport implements CodexAppServerTransport { #providerIdentity: Record | null = null; #turnControls: NativeTurnControlCapabilities = { steering: false, queuedFollowUp: false }; + supportsTurnReasoning(): boolean { + if (this.options.provider !== "opencode") return false; + // Use the durable selected model on both fresh and recovered sessions. + // Older states may retain the provider only in the preparation command. + const state = this.#core?.store.state; + const template = this.#runAttachTemplate ?? state?.runAttachTemplate + ?? state?.commands.findLast(command => + (command.type === "run.prepare" || command.type === "run.attach") + && record(command.payload).provider !== undefined, + )?.payload; + const model = record(record(template).provider).model; + return typeof model === "string" && model.startsWith("openrouter/"); + } + turnControlCapabilities(): NativeTurnControlCapabilities | null { if (this.options.provider !== "acpx") return null; if (this.#closed || this.#failure) return { steering: false, queuedFollowUp: false }; @@ -5883,6 +5897,11 @@ class DurablePrpCodexTransport implements CodexAppServerTransport { .filter((item) => item.type !== "skill") .map((item) => (typeof item.text === "string" ? item.text : "")) .join("\n"); + const reasoningMode = params.reasoningMode; + if (reasoningMode !== undefined && ( + !this.supportsTurnReasoning() + || (reasoningMode !== "default" && reasoningMode !== "disabled") + )) throw new Error("turn.reasoningMode requires an OpenCode/OpenRouter model and must be default or disabled"); const skills = resolveRunnerdCodexSkillInputs( input.filter((item) => item.type === "skill"), this.options.runtimeContext ?? null, @@ -5916,6 +5935,7 @@ class DurablePrpCodexTransport implements CodexAppServerTransport { "turn.start", { text: message, + ...(reasoningMode === undefined ? {} : { reasoningMode }), ...(skills.length ? { skills } : {}), turnId: pendingTurnId, }, diff --git a/packages/paperclip-runner/test/fixtures/fake-opencode-server.mjs b/packages/paperclip-runner/test/fixtures/fake-opencode-server.mjs index 69bd35a8e4e..692fd9c2ebd 100644 --- a/packages/paperclip-runner/test/fixtures/fake-opencode-server.mjs +++ b/packages/paperclip-runner/test/fixtures/fake-opencode-server.mjs @@ -99,6 +99,8 @@ await writeFile( home: process.env.HOME, configHome: process.env.XDG_CONFIG_HOME, projectConfigDisabled: process.env.OPENCODE_DISABLE_PROJECT_CONFIG, + modelsFetchDisabled: process.env.OPENCODE_DISABLE_MODELS_FETCH, + defaultPluginsDisabled: process.env.OPENCODE_DISABLE_DEFAULT_PLUGINS, }), ); @@ -400,8 +402,7 @@ const server = createServer(async (request, response) => { const promptPayload = JSON.parse(Buffer.concat(chunks).toString("utf8")); await appendFile(join(process.env.XDG_DATA_HOME, "fake-prompt-requests.ndjson"), JSON.stringify(promptPayload) + "\n"); if ( - promptPayload.providerID !== "openrouter" || - promptPayload.modelID !== "deepseek/deepseek-v4-flash-0731" || + `${promptPayload.providerID}/${promptPayload.modelID}` !== runtimeConfig.model || "model" in promptPayload ) { return json(response, 400, { @@ -427,11 +428,56 @@ const server = createServer(async (request, response) => { }, }, }); + emit({ + type: "message.part.delta", + properties: { + sessionID: session.id, + messageID: "message-late-source", + partID: "part-late-straggler", + field: "text", + delta: "late straggler text must not reach the next turn", + }, + }); } } setTimeout(async () => { await callFirstPaperclipTool(); const parsedPrompt = parsedPromptText(promptPayload); + if (String(parsedPrompt.message ?? "").includes("incremental-stream")) { + const messageID = `stream-message-${promptTurnSeq}`; + const part = { id: `stream-part-${promptTurnSeq}`, messageID, sessionID: session.id, type: "text", text: "" }; + const reasoning = { ...part, id: `${part.id}-reasoning`, type: "reasoning" }; + const update = (value) => emit({ type: "message.part.updated", properties: { sessionID: session.id, part: value } }); + const delta = (value, text, extra = {}) => emit({ type: "message.part.delta", properties: { sessionID: session.id, messageID, partID: value.id, field: "text", delta: text, ...extra } }); + const announceRole = () => emit({ type: "message.updated", properties: { sessionID: session.id, info: { id: messageID, role: "assistant" } } }); + const roleFirst = String(parsedPrompt.message).includes("role-first"); + const longResponse = String(parsedPrompt.message).includes("role-last-long"); + if (roleFirst) announceRole(); + // Parts and deltas can precede the message's role announcement. + update(part); + update(reasoning); + delta(reasoning, "Thinking"); + delta(part, "Hi "); + const emojiCount = longResponse ? 120 : 2; + for (let i = 0; i < emojiCount; i++) delta(part, "👋"); + delta(part, "wrong message", { messageID: "unrelated-message" }); + delta(part, "wrong session", { sessionID: "unrelated-session" }); + delta(part, "wrong field", { field: "output" }); + delta(part, "unknown part", { partID: "unknown-part" }); + const finalPart = { ...part, text: `Hi ${"👋".repeat(emojiCount)}`, time: { start: 1, end: 2 } }; + if (longResponse) update(finalPart); + if (!roleFirst) announceRole(); + // The consumer must observe a delta before the completed snapshot + // exists. This handshake makes the streaming regression deterministic. + for (let i = 0; i < 500; i++) { + if (await readFile(join(process.env.XDG_DATA_HOME, "release-stream"), "utf8").catch(() => "") === "ready") break; + await new Promise(resolve => setTimeout(resolve, 10)); + } + await writeFile(join(process.env.XDG_DATA_HOME, "stream-released"), "ready"); + update(finalPart); + emit({ type: "session.idle", properties: { sessionID: session.id } }); + return; + } if (String(parsedPrompt.message ?? "").includes("native-question")) { pendingQuestion = nativeQuestion(); emit({ diff --git a/server/src/__tests__/heartbeat-responsible-user-invariant.test.ts b/server/src/__tests__/heartbeat-responsible-user-invariant.test.ts index 59c3cb2c419..54bcebe2005 100644 --- a/server/src/__tests__/heartbeat-responsible-user-invariant.test.ts +++ b/server/src/__tests__/heartbeat-responsible-user-invariant.test.ts @@ -23,6 +23,7 @@ import { import { heartbeatService } from "../services/heartbeat.ts"; import { runningProcesses } from "../adapters/index.ts"; import { drainHeartbeatRunsToQuiescence } from "./helpers/drain-heartbeat-runs.js"; +import { createLifecycleDriver } from "../services/agent-lifecycle-driver.js"; const mockAdapterExecute = vi.hoisted(() => vi.fn(async () => ({ @@ -159,6 +160,27 @@ describeEmbeddedPostgres("heartbeat responsible-user invariant", () => { return { companyId, ownerUserId, agentId }; } + it("dispatches saved work on readiness with company scoping and normal admission checks", async () => { + const { companyId, agentId, ownerUserId } = await seedCompany(); + await db.update(agents).set({ status: "paused", lifecycleState: "verifying" }).where(eq(agents.id, agentId)); + const [run] = await db.insert(heartbeatRuns).values({ companyId, agentId, status: "queued", + invocationSource: "on_demand", contextSnapshot: { responsibleUserId: ownerUserId } }).returning(); + const driver = createLifecycleDriver(db, {} as never); + const currentAgent = async () => (await db.select().from(agents).where(eq(agents.id, agentId)))[0]; + await driver.onReady!(await currentAgent()); + expect(mockAdapterExecute).not.toHaveBeenCalled(); + expect((await db.select().from(heartbeatRuns).where(eq(heartbeatRuns.id, run.id)))[0].status).toBe("queued"); + + await db.update(agents).set({ status: "idle", lifecycleState: "ready" }).where(eq(agents.id, agentId)); + await heartbeat.resumeQueuedRunsForAgent(randomUUID(), agentId); + expect(mockAdapterExecute).not.toHaveBeenCalled(); + expect((await db.select().from(heartbeatRuns).where(eq(heartbeatRuns.id, run.id)))[0].status).toBe("queued"); + + await driver.onReady!(await currentAgent()); + expect(await waitForRun(db, run.id)).toMatchObject({ status: "succeeded", responsibleUserId: ownerUserId }); + expect(mockAdapterExecute).toHaveBeenCalledOnce(); + }); + it("dispatches an interrupted queue under the clicking operator through the real startup path", async () => { const { companyId, agentId, ownerUserId } = await seedCompany(); const operatorId = `operator-${randomUUID()}`, issueId = randomUUID(), commentId = randomUUID(), queueId = randomUUID(); diff --git a/server/src/modules/agent-lifecycle/__tests__/diagnostics.test.ts b/server/src/modules/agent-lifecycle/__tests__/diagnostics.test.ts index 4f7a8db7214..06b7e246d49 100644 --- a/server/src/modules/agent-lifecycle/__tests__/diagnostics.test.ts +++ b/server/src/modules/agent-lifecycle/__tests__/diagnostics.test.ts @@ -13,14 +13,58 @@ function fixture() { const store = { get: vi.fn(), claim: vi.fn().mockResolvedValue(agent), renew: vi.fn().mockResolvedValue(true), setRequiredPlugins: vi.fn().mockResolvedValue(true), completeHost: vi.fn().mockResolvedValue(true), completePlugin: vi.fn().mockResolvedValue(true), defer: vi.fn().mockResolvedValue(true), - change: vi.fn().mockResolvedValue({ ...agent, lifecycleState: "ready" }), pending: vi.fn().mockResolvedValue([agent.id]) } satisfies LifecycleStore; + change: vi.fn().mockResolvedValue({ ...agent, lifecycleState: "ready", lifecycleVersion: 4, lifecycleOperation: null }), pending: vi.fn().mockResolvedValue([agent.id]) } satisfies LifecycleStore; const driver = { requiredPluginIds: vi.fn().mockResolvedValue(["plugin"]), runHost: vi.fn().mockResolvedValue("complete"), - runPlugin: vi.fn().mockResolvedValue("complete") } satisfies LifecycleDriver; + runPlugin: vi.fn().mockResolvedValue("complete"), onReady: vi.fn().mockResolvedValue(undefined) } satisfies LifecycleDriver; const report = vi.fn(); return { store, driver, report, worker: createLifecycleWorker(store, driver, report) }; } afterEach(() => { vi.restoreAllMocks(); vi.useRealTimers(); }); +describe("lifecycle readiness handoff", () => { + it("notifies the scheduler only after host, plugins, and readiness have committed", async () => { + const { store, driver, worker } = fixture(); + let commit!: (ready: LifecycleAgent) => void; + store.change.mockReturnValue(new Promise(resolve => { commit = resolve; })); + driver.onReady.mockImplementation(async ready => { + expect(store.completeHost).toHaveBeenCalledOnce(); + expect(store.completePlugin).toHaveBeenCalledOnce(); + expect(store.change).toHaveBeenCalledOnce(); + expect(ready).toMatchObject({ lifecycleState: "ready", lifecycleVersion: 4, lifecycleOperation: null }); + }); + const work = worker.process(agent.id); + await vi.waitFor(() => expect(store.change).toHaveBeenCalledOnce()); + expect(driver.onReady).not.toHaveBeenCalled(); + commit({ ...agent, lifecycleState: "ready", lifecycleVersion: 4, lifecycleOperation: null }); + await work; + expect(driver.onReady).toHaveBeenCalledOnce(); + await worker.stop(); + }); + + it.each(["host", "plugin", "transition", "hold"])("does not dispatch when %s prevents readiness", async stage => { + const { store, driver, worker } = fixture(); + if (stage === "host") driver.runHost.mockResolvedValue("pending"); + if (stage === "plugin") driver.runPlugin.mockResolvedValue("pending"); + if (stage === "transition") store.change.mockResolvedValue(null); + if (stage === "hold") store.change.mockResolvedValue({ ...agent, lifecycleState: "paused" }); + await worker.process(agent.id); + expect(driver.onReady).not.toHaveBeenCalled(); + await worker.stop(); + }); + + it("preserves committed readiness after a dispatch failure for the queue sweep to recover", async () => { + const { store, driver, report, worker } = fixture(); + const error = new Error("queue unavailable"); + driver.onReady.mockRejectedValue(error); + await worker.process(agent.id); + expect(report).toHaveBeenCalledWith(expect.objectContaining({ stage: "ready", phase: "ready", version: 4 }), error); + expect(store.defer).not.toHaveBeenCalled(); + expect(driver.runHost).toHaveBeenCalledOnce(); + expect(store.change).toHaveBeenCalledOnce(); + await worker.stop(); + }); +}); + describe("lifecycle failure diagnostics", () => { it.each(["host", "plugin", "claim", "renew", "scan", "defer"] as const)("reports %s failures with the available context", async stage => { const { store, driver, report, worker } = fixture(); diff --git a/server/src/modules/agent-lifecycle/application/ports.ts b/server/src/modules/agent-lifecycle/application/ports.ts index 1845ad8b7d1..11442b70b88 100644 --- a/server/src/modules/agent-lifecycle/application/ports.ts +++ b/server/src/modules/agent-lifecycle/application/ports.ts @@ -27,11 +27,13 @@ export interface LifecycleDriver { requiredPluginIds(agent: LifecycleAgent): Promise; runHost(agent: LifecycleAgent): Promise<"complete" | "pending">; runPlugin(agent: LifecycleAgent, pluginId: string): Promise<"complete" | "pending">; + /** Notify the scheduler only after readiness has committed. */ + onReady?(agent: LifecycleAgent): Promise; } export interface LifecycleFailure { stage: "claim" | "renew" | "select_plugins" | "save_plugins" | "host" | "plugin" - | "complete_host" | "complete_plugin" | "transition" | "defer" | "scan" | "policy_scan"; + | "complete_host" | "complete_plugin" | "transition" | "defer" | "scan" | "policy_scan" | "ready"; agentId?: string; companyId?: string; phase?: AgentLifecycleState; diff --git a/server/src/modules/agent-lifecycle/application/worker.ts b/server/src/modules/agent-lifecycle/application/worker.ts index ad1c143dd96..3b90b227673 100644 --- a/server/src/modules/agent-lifecycle/application/worker.ts +++ b/server/src/modules/agent-lifecycle/application/worker.ts @@ -56,6 +56,14 @@ export function createLifecycleWorker(store: LifecycleStore, driver: LifecycleDr if (!await runStep("complete_plugin", () => store.completePlugin(agent, owner, pluginId), pluginId)) continue phase; } const changed = await runStep("transition", () => store.change(id, "complete", { version: agent.lifecycleVersion, owner })); + if (changed?.lifecycleState === "ready" && !stopped && canRun()) { + // Accepted messages may already be waiting for verification. Do not + // leave them until the periodic queue sweep. Readiness is committed; + // a dispatch failure must not replay setup or defer the old phase. + await attempt({ ...context, stage: "ready", phase: changed.lifecycleState, + version: changed.lifecycleVersion, operationId: changed.lifecycleOperation?.id }, + async () => { await driver.onReady?.(changed); }).catch(() => {}); + } if (changed && !isLifecycleWorkPending(changed.lifecycleState)) return; } catch { if (!await runStep("defer", () => store.defer(agent, owner, "The lifecycle step failed. Retry the operation.", new Date()))) continue; diff --git a/server/src/services/agent-lifecycle-driver.ts b/server/src/services/agent-lifecycle-driver.ts index 68cbea29999..11963d997fa 100644 --- a/server/src/services/agent-lifecycle-driver.ts +++ b/server/src/services/agent-lifecycle-driver.ts @@ -10,6 +10,9 @@ export function createLifecycleDriver(db: Db, manager: PluginWorkerManager): Lif const { verify } = agentHarnessVerificationService(db, manager); const heartbeat = heartbeatService(db, { pluginWorkerManager: manager }); return { + async onReady(agent) { + await heartbeat.resumeQueuedRunsForAgent(agent.companyId, agent.id); + }, async requiredPluginIds(agent) { if (agent.lifecycleRequiredPluginIds !== null) return agent.lifecycleRequiredPluginIds; const installed = await db.select().from(plugins).where(inArray(plugins.status, ["ready", "error", "installed", "upgrade_pending"])); diff --git a/server/src/services/heartbeat.ts b/server/src/services/heartbeat.ts index a549ee7ce42..961019eca34 100644 --- a/server/src/services/heartbeat.ts +++ b/server/src/services/heartbeat.ts @@ -8088,6 +8088,12 @@ export function heartbeatService( resumeQueuedRuns, + resumeQueuedRunsForAgent: async (companyId: string, agentId: string) => { + const agent = await getAgent(agentId); + if (!agent || agent.companyId !== companyId) return []; + return startNextQueuedRunForAgent(agentId); + }, + scheduleBoundedRetry, reconcileStrandedAssignedIssues,