@poncho-ai/harness 0.14.0 → 0.14.2

package/.turbo/turbo-build.log

@@ -1,5 +1,5 @@
 
-> @poncho-ai/harness@0.14.0 build /home/runner/work/poncho-ai/poncho-ai/packages/harness
+> @poncho-ai/harness@0.14.2 build /Users/cesar/Dev/latitude/poncho-ai/packages/harness
 > tsup src/index.ts --format esm --dts
 
 CLI Building entry: src/index.ts
@@ -7,8 +7,8 @@
 CLI tsup v8.5.1
 CLI Target: es2022
 ESM Build start
-ESM dist/index.js 173.62 KB
-ESM ⚡️ Build success in 98ms
+ESM dist/index.js 177.01 KB
+ESM ⚡️ Build success in 73ms
 DTS Build start
-DTS ⚡️ Build success in 5281ms
-DTS dist/index.d.ts 20.92 KB
+DTS ⚡️ Build success in 3459ms
+DTS dist/index.d.ts 21.71 KB

package/.turbo/turbo-lint.log

@@ -0,0 +1,6 @@
+
+> @poncho-ai/harness@0.11.2 lint /Users/cesar/Dev/latitude/poncho-ai/packages/harness
+> eslint src/
+
+sh: eslint: command not found
+ ELIFECYCLE  Command failed.

package/.turbo/turbo-test.log

@@ -0,0 +1,137 @@
+
+> @poncho-ai/harness@0.14.0 test /Users/cesar/Dev/latitude/poncho-ai/packages/harness
+> vitest
+
+
+ RUN  v1.6.1 /Users/cesar/Dev/latitude/poncho-ai/packages/harness
+
+[event] step:completed {"type":"step:completed","step":1,"duration":1}
+[event] step:started {"type":"step:started","step":2}
+ ✓ test/telemetry.test.ts  (3 tests) 11ms
+ ✓ test/schema-converter.test.ts  (27 tests) 21ms
+stdout | test/mcp.test.ts > mcp bridge protocol transports > discovers and calls tools over streamable HTTP
+[poncho][mcp] {"event":"catalog.loaded","server":"remote","discoveredCount":1}
+[poncho][mcp] {"event":"tools.selected","requestedPatternCount":1,"registeredCount":1,"filteredByPolicyCount":0,"filteredByIntentCount":0}
+
+stdout | test/mcp.test.ts > mcp bridge protocol transports > selects discovered tools by requested patterns
+[poncho][mcp] {"event":"catalog.loaded","server":"remote","discoveredCount":2}
+[poncho][mcp] {"event":"tools.selected","requestedPatternCount":1,"registeredCount":1,"filteredByPolicyCount":0,"filteredByIntentCount":1}
+[poncho][mcp] {"event":"tools.selected","requestedPatternCount":1,"registeredCount":2,"filteredByPolicyCount":0,"filteredByIntentCount":0}
+
+stdout | test/mcp.test.ts > mcp bridge protocol transports > skips discovery when bearer token env value is missing
+[poncho][mcp] {"event":"tools.selected","requestedPatternCount":1,"registeredCount":0,"filteredByPolicyCount":0,"filteredByIntentCount":0}
+
+stderr | test/mcp.test.ts > mcp bridge protocol transports > skips discovery when bearer token env value is missing
+[poncho][mcp] {"event":"auth.token_missing","server":"remote","tokenEnv":"MISSING_TOKEN_ENV"}
+
+stdout | test/mcp.test.ts > mcp bridge protocol transports > returns actionable errors for 403 permission failures
+[poncho][mcp] {"event":"catalog.loaded","server":"remote","discoveredCount":1}
+[poncho][mcp] {"event":"tools.selected","requestedPatternCount":1,"registeredCount":1,"filteredByPolicyCount":0,"filteredByIntentCount":0}
+
+ ✓ test/mcp.test.ts  (6 tests) 516ms
+ ✓ test/agent-parser.test.ts  (10 tests) 130ms
+ ✓ test/memory.test.ts  (4 tests) 192ms
+ ✓ test/state.test.ts  (5 tests) 294ms
+ ✓ test/model-factory.test.ts  (4 tests) 8ms
+ ✓ test/agent-identity.test.ts  (2 tests) 33ms
+stdout | test/harness.test.ts > agent harness > registers default filesystem tools
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > disables write_file by default in production environment
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > allows disabling built-in tools via poncho.config.js
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > supports per-environment tool overrides
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > supports per-environment tool overrides
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > does not auto-register exported tool objects from skill scripts
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > refreshes skill metadata and tools in development mode
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > refreshes skill metadata and tools in development mode
+[poncho][mcp] {"event":"tools.cleared","reason":"skills:changed","requestedPatterns":[]}
+[poncho][mcp] {"event":"tools.cleared","reason":"activate:beta","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > prunes removed active skills after refresh in development mode
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+[poncho][mcp] {"event":"tools.cleared","reason":"activate:obsolete","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > prunes removed active skills after refresh in development mode
+[poncho][mcp] {"event":"tools.cleared","reason":"skills:changed","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > does not refresh skills outside development mode
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > clears active skills when skill metadata changes in development mode
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > clears active skills when skill metadata changes in development mode
+[poncho][mcp] {"event":"tools.cleared","reason":"activate:alpha","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > clears active skills when skill metadata changes in development mode
+[poncho][mcp] {"event":"tools.cleared","reason":"skills:changed","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > lists skill scripts through list_skill_scripts
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > runs JavaScript/TypeScript skill scripts through run_skill_script
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > runs AGENT-scope scripts from root scripts directory
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > blocks path traversal in run_skill_script
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > requires allowed-tools entries for non-standard script directories
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > registers MCP tools dynamically for stacked active skills and supports deactivation
+[poncho][mcp] {"event":"catalog.loaded","server":"remote","discoveredCount":2}
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+[poncho][mcp] {"event":"tools.selected","requestedPatternCount":1,"registeredCount":1,"filteredByPolicyCount":0,"filteredByIntentCount":1}
+[poncho][mcp] {"event":"tools.refreshed","reason":"activate:skill-a","requestedPatterns":["remote/a"],"registeredCount":1,"activeSkills":["skill-a"]}
+[poncho][mcp] {"event":"tools.selected","requestedPatternCount":2,"registeredCount":2,"filteredByPolicyCount":0,"filteredByIntentCount":0}
+[poncho][mcp] {"event":"tools.refreshed","reason":"activate:skill-b","requestedPatterns":["remote/a","remote/b"],"registeredCount":2,"activeSkills":["skill-a","skill-b"]}
+[poncho][mcp] {"event":"tools.selected","requestedPatternCount":1,"registeredCount":1,"filteredByPolicyCount":0,"filteredByIntentCount":1}
+[poncho][mcp] {"event":"tools.refreshed","reason":"deactivate:skill-a","requestedPatterns":["remote/b"],"registeredCount":1,"activeSkills":["skill-b"]}
+
+stdout | test/harness.test.ts > agent harness > supports flat tool access config format
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > flat tool access takes priority over legacy defaults
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > byEnvironment overrides flat tool access
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > registerTools skips tools disabled via config
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > approval access level registers the tool but marks it for approval
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > tools without approval config do not require approval
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+
+stdout | test/harness.test.ts > agent harness > allows in-flight MCP calls to finish after skill deactivation
+[poncho][mcp] {"event":"catalog.loaded","server":"remote","discoveredCount":1}
+[poncho][mcp] {"event":"tools.cleared","reason":"initialize","requestedPatterns":[]}
+[poncho][mcp] {"event":"tools.selected","requestedPatternCount":1,"registeredCount":1,"filteredByPolicyCount":0,"filteredByIntentCount":0}
+[poncho][mcp] {"event":"tools.refreshed","reason":"activate:skill-slow","requestedPatterns":["remote/slow"],"registeredCount":1,"activeSkills":["skill-slow"]}
+[poncho][mcp] {"event":"tools.cleared","reason":"deactivate:skill-slow","requestedPatterns":[]}
+
+ ✓ test/harness.test.ts  (25 tests) 429ms
+
+ Test Files  9 passed (9)
+      Tests  86 passed (86)
+   Start at  10:02:54
+   Duration  4.19s (transform 2.16s, setup 0ms, collect 7.51s, tests 1.63s, environment 1ms, prepare 1.97s)
+

package/CHANGELOG.md

@@ -1,5 +1,31 @@
 # @poncho-ai/harness
 
+## 0.14.2
+
+### Patch Changes
+
+- [`1f47bb4`](https://github.com/cesr/poncho-ai/commit/1f47bb49e5d48dc17644172012b057190b316469) Thanks [@cesr](https://github.com/cesr)! - Add conversation rename via double-click on the title in the web UI, standardize all credential config fields to the `*Env` naming pattern, and sync the init README template with the repo README.
+
+- Updated dependencies [[`1f47bb4`](https://github.com/cesr/poncho-ai/commit/1f47bb49e5d48dc17644172012b057190b316469)]:
+  - @poncho-ai/sdk@1.0.3
+
+## 0.14.1
+
+### Patch Changes
+
+- [`e000b96`](https://github.com/cesr/poncho-ai/commit/e000b96837cbbb8d95c868c91a614f458868c444) Thanks [@cesr](https://github.com/cesr)! - Durable approval checkpoints, email conversation improvements, and web UI fixes
+  - Simplify approval system to checkpoint-only (remove legacy blocking approvalHandler)
+  - Optimize checkpoint storage with delta messages instead of full history
+  - Add sidebar sections for conversations awaiting approval with status indicator
+  - Fix nested checkpoint missing baseMessageCount in resumeRunFromCheckpoint
+  - Improve email conversation titles (sender email + subject)
+  - Remove email threading — each incoming email creates its own conversation
+  - Fix streaming after approval to preserve existing messages (liveOnly mode)
+  - Preserve newlines in user messages in web UI
+
+- Updated dependencies [[`e000b96`](https://github.com/cesr/poncho-ai/commit/e000b96837cbbb8d95c868c91a614f458868c444)]:
+  - @poncho-ai/sdk@1.0.2
+
 ## 0.14.0
 
 ### Minor Changes

package/dist/index.d.ts

@@ -1,4 +1,4 @@
-import { Message, ToolDefinition, RunInput, AgentEvent, RunResult, JsonSchema, ToolContext } from '@poncho-ai/sdk';
+import { Message, ToolDefinition, ToolContext, RunInput, AgentEvent, RunResult, JsonSchema } from '@poncho-ai/sdk';
 export { ToolDefinition, defineTool } from '@poncho-ai/sdk';
 import { LanguageModel } from 'ai';
 import { z } from 'zod';
@@ -84,7 +84,15 @@ interface Conversation {
         approvalId: string;
         runId: string;
         tool: string;
+        toolCallId?: string;
         input: Record<string, unknown>;
+        checkpointMessages?: Message[];
+        baseMessageCount?: number;
+        pendingToolCalls?: Array<{
+            id: string;
+            name: string;
+            input: Record<string, unknown>;
+        }>;
     }>;
     ownerId: string;
     tenantId: string | null;
@@ -103,8 +111,8 @@ type StateProviderName = "local" | "memory" | "redis" | "upstash" | "dynamodb";
 interface StateConfig {
     provider?: StateProviderName;
     ttl?: number;
-    url?: string;
-    token?: string;
+    urlEnv?: string;
+    tokenEnv?: string;
     table?: string;
     region?: string;
 }
@@ -146,8 +154,8 @@ interface MainMemory {
 interface MemoryConfig {
     enabled?: boolean;
     provider?: StateProviderName;
-    url?: string;
-    token?: string;
+    urlEnv?: string;
+    tokenEnv?: string;
     table?: string;
     region?: string;
     ttl?: number;
@@ -210,8 +218,8 @@ declare class LocalMcpBridge {
 
 interface StorageConfig {
     provider?: "local" | "memory" | "redis" | "upstash" | "dynamodb";
-    url?: string;
-    token?: string;
+    urlEnv?: string;
+    tokenEnv?: string;
     table?: string;
     region?: string;
     ttl?: number | {
@@ -266,6 +274,7 @@ interface PonchoConfig extends McpConfig {
         required?: boolean;
         type?: "bearer" | "header" | "custom";
         headerName?: string;
+        tokenEnv?: string;
         validate?: (token: string, req?: unknown) => Promise<boolean> | boolean;
     };
     state?: {
@@ -275,12 +284,20 @@ interface PonchoConfig extends McpConfig {
     };
     memory?: MemoryConfig;
     storage?: StorageConfig;
+    providers?: {
+        openai?: {
+            apiKeyEnv?: string;
+        };
+        anthropic?: {
+            apiKeyEnv?: string;
+        };
+    };
     telemetry?: {
         enabled?: boolean;
         otlp?: string;
         latitude?: {
-            apiKey?: string;
-            projectId?: string | number;
+            apiKeyEnv?: string;
+            projectIdEnv?: string;
             path?: string;
             documentPath?: string;
         };
@@ -355,24 +372,48 @@ type ModelProviderFactory = (modelName: string) => LanguageModel;
  * resolve via the base prefix. Longest match wins.
  */
 declare const getModelContextWindow: (modelName: string) => number;
+interface ProviderConfig {
+    openai?: {
+        apiKeyEnv?: string;
+    };
+    anthropic?: {
+        apiKeyEnv?: string;
+    };
+}
 /**
- * Creates a model provider factory for the specified AI provider
- * @param provider - The provider name ('openai' or 'anthropic')
- * @returns A function that takes a model name and returns a LanguageModel instance
+ * Creates a model provider factory for the specified AI provider.
+ * API keys are read from environment variables; override the env var
+ * name via the `providers` config in `poncho.config.js`.
  */
-declare const createModelProvider: (provider?: string) => ModelProviderFactory;
+declare const createModelProvider: (provider?: string, config?: ProviderConfig) => ModelProviderFactory;
+
+interface ToolCall {
+    id: string;
+    name: string;
+    input: Record<string, unknown>;
+}
+interface ToolExecutionResult {
+    callId: string;
+    tool: string;
+    output?: unknown;
+    error?: string;
+}
+declare class ToolDispatcher {
+    private readonly tools;
+    register(tool: ToolDefinition): void;
+    registerMany(tools: ToolDefinition[]): void;
+    unregister(name: string): void;
+    unregisterMany(names: Iterable<string>): void;
+    list(): ToolDefinition[];
+    get(name: string): ToolDefinition | undefined;
+    execute(call: ToolCall, context: ToolContext): Promise<ToolExecutionResult>;
+    executeBatch(calls: ToolCall[], context: ToolContext): Promise<ToolExecutionResult[]>;
+}
 
 interface HarnessOptions {
     workingDir?: string;
     environment?: "development" | "staging" | "production";
     toolDefinitions?: ToolDefinition[];
-    approvalHandler?: (request: {
-        tool: string;
-        input: Record<string, unknown>;
-        runId: string;
-        step: number;
-        approvalId: string;
-    }) => Promise<boolean> | boolean;
     modelProvider?: ModelProviderFactory;
     uploadStore?: UploadStore;
 }
@@ -388,7 +429,6 @@ declare class AgentHarness {
     private modelProvider;
     private readonly modelProviderInjected;
     private readonly dispatcher;
-    private readonly approvalHandler?;
     readonly uploadStore?: UploadStore;
     private skillContextWindow;
     private memoryStore?;
@@ -439,6 +479,19 @@ declare class AgentHarness {
      */
     runWithTelemetry(input: RunInput): AsyncGenerator<AgentEvent>;
     run(input: RunInput): AsyncGenerator<AgentEvent>;
+    executeTools(calls: ToolCall[], context: ToolContext): Promise<ToolExecutionResult[]>;
+    continueFromToolResult(input: {
+        messages: Message[];
+        toolResults: Array<{
+            callId: string;
+            toolName: string;
+            result?: unknown;
+            error?: string;
+        }>;
+        conversationId?: string;
+        parameters?: Record<string, unknown>;
+        abortSignal?: AbortSignal;
+    }): AsyncGenerator<AgentEvent>;
     runToCompletion(input: RunInput): Promise<HarnessRunOutput>;
 }
 
@@ -453,8 +506,8 @@ declare class AgentHarness {
  * Vercel AI SDK support.
  */
 interface LatitudeCaptureConfig {
-    apiKey?: string;
-    projectId?: string | number;
+    apiKeyEnv?: string;
+    projectIdEnv?: string;
     path?: string;
     defaultPath?: string;
 }
@@ -549,8 +602,8 @@ interface TelemetryConfig {
     enabled?: boolean;
     otlp?: string;
     latitude?: {
-        apiKey?: string;
-        projectId?: string | number;
+        apiKeyEnv?: string;
+        projectIdEnv?: string;
         path?: string;
         documentPath?: string;
     };
@@ -563,27 +616,4 @@ declare class TelemetryEmitter {
     private sendOtlp;
 }
 
-interface ToolCall {
-    id: string;
-    name: string;
-    input: Record<string, unknown>;
-}
-interface ToolExecutionResult {
-    callId: string;
-    tool: string;
-    output?: unknown;
-    error?: string;
-}
-declare class ToolDispatcher {
-    private readonly tools;
-    register(tool: ToolDefinition): void;
-    registerMany(tools: ToolDefinition[]): void;
-    unregister(name: string): void;
-    unregisterMany(names: Iterable<string>): void;
-    list(): ToolDefinition[];
-    get(name: string): ToolDefinition | undefined;
-    execute(call: ToolCall, context: ToolContext): Promise<ToolExecutionResult>;
-    executeBatch(calls: ToolCall[], context: ToolContext): Promise<ToolExecutionResult[]>;
-}
-
-export { type AgentFrontmatter, AgentHarness, type AgentIdentity, type AgentLimitsConfig, type AgentModelConfig, type BuiltInToolToggles, type Conversation, type ConversationState, type ConversationStore, type CronJobConfig, type HarnessOptions, type HarnessRunOutput, InMemoryConversationStore, InMemoryStateStore, LatitudeCapture, type LatitudeCaptureConfig, LocalMcpBridge, LocalUploadStore, type MainMemory, type McpConfig, type MemoryConfig, type MemoryStore, type MessagingChannelConfig, type ModelProviderFactory, PONCHO_UPLOAD_SCHEME, type ParsedAgent, type PonchoConfig, type RemoteMcpServerConfig, type RuntimeRenderContext, S3UploadStore, STORAGE_SCHEMA_VERSION, type SkillContextEntry, type SkillMetadata, type StateConfig, type StateProviderName, type StateStore, type StorageConfig, type TelemetryConfig, TelemetryEmitter, type ToolAccess, type ToolCall, ToolDispatcher, type ToolExecutionResult, type UploadStore, type UploadsConfig, VercelBlobUploadStore, buildAgentDirectoryName, buildSkillContextWindow, createConversationStore, createDefaultTools, createMemoryStore, createMemoryTools, createModelProvider, createSkillTools, createStateStore, createUploadStore, createWriteTool, deriveUploadKey, ensureAgentIdentity, generateAgentId, getAgentStoreDirectory, getModelContextWindow, getPonchoStoreRoot, jsonSchemaToZod, loadPonchoConfig, loadSkillContext, loadSkillInstructions, loadSkillMetadata, normalizeScriptPolicyPath, parseAgentFile, parseAgentMarkdown, readSkillResource, renderAgentPrompt, resolveAgentIdentity, resolveMemoryConfig, resolveSkillDirs, resolveStateConfig, slugifyStorageComponent };
+export { type AgentFrontmatter, AgentHarness, type AgentIdentity, type AgentLimitsConfig, type AgentModelConfig, type BuiltInToolToggles, type Conversation, type ConversationState, type ConversationStore, type CronJobConfig, type HarnessOptions, type HarnessRunOutput, InMemoryConversationStore, InMemoryStateStore, LatitudeCapture, type LatitudeCaptureConfig, LocalMcpBridge, LocalUploadStore, type MainMemory, type McpConfig, type MemoryConfig, type MemoryStore, type MessagingChannelConfig, type ModelProviderFactory, PONCHO_UPLOAD_SCHEME, type ParsedAgent, type PonchoConfig, type ProviderConfig, type RemoteMcpServerConfig, type RuntimeRenderContext, S3UploadStore, STORAGE_SCHEMA_VERSION, type SkillContextEntry, type SkillMetadata, type StateConfig, type StateProviderName, type StateStore, type StorageConfig, type TelemetryConfig, TelemetryEmitter, type ToolAccess, type ToolCall, ToolDispatcher, type ToolExecutionResult, type UploadStore, type UploadsConfig, VercelBlobUploadStore, buildAgentDirectoryName, buildSkillContextWindow, createConversationStore, createDefaultTools, createMemoryStore, createMemoryTools, createModelProvider, createSkillTools, createStateStore, createUploadStore, createWriteTool, deriveUploadKey, ensureAgentIdentity, generateAgentId, getAgentStoreDirectory, getModelContextWindow, getPonchoStoreRoot, jsonSchemaToZod, loadPonchoConfig, loadSkillContext, loadSkillInstructions, loadSkillMetadata, normalizeScriptPolicyPath, parseAgentFile, parseAgentMarkdown, readSkillResource, renderAgentPrompt, resolveAgentIdentity, resolveMemoryConfig, resolveSkillDirs, resolveStateConfig, slugifyStorageComponent };