newo 3.7.3 → 3.7.5

package/CHANGELOG.md

@@ -7,6 +7,37 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [3.7.5] - 2026-06-17
+
+### Added
+
+- **`newo sandbox` connector selection (R1)** — new `--connector <connector_idn>` and `--integration <idn>` flags select which connector to chat through instead of always taking the first running connector of the `sandbox` integration (which made agents like Vibe Builder's `vibe_agent` unreachable from the CLI when another connector, e.g. `convo_agent_sandbox`, came first). `--list-connectors` prints the running connectors (with `--json` for machine output). When `--connector` is not found, the error lists the available running connectors. Without flags, behavior is unchanged (first running connector). Surfaced during PR #2076 superagent testing.
+- **`newo sandbox` long messages and machine-readable output (R2)** — `--file <path>` and `--stdin` send message text without shell-argument limits (tested scenarios used 40k–400k char messages); `--timeout <seconds>` overrides the previously hardcoded 60-second response-poll limit (long VibeFlow turns ran 1–7 minutes); `--json` emits `{actor_id, persona_id, connector_idn, external_event_id, user_external_event_id, agent_external_event_id, response, elapsed_ms, timed_out, flow_idn, skill_idn, session_id}`. The `external_event_id` of the user turn is the correlation key for `newo logs --event-id`, closing the gap that previously required ad-hoc scripts importing `dist/api.js` directly.
+- **`newo get-skill` / `newo update-skill` (R3)** — point inspection and point edits of a single skill on the platform by IDN path (`<skill-idn> --project <idn> --agent <idn> --flow <idn>`), with no pulled workspace required. `get-skill` shows the live model, runner_type, parameters, and prompt_script (`--json` supported). `update-skill` changes `--model <provider_idn>/<model_idn>` and/or `--script <file>`, leaving everything else untouched; `--publish` (with optional `--publish-description`) publishes the flow afterwards, otherwise changes stay draft. Warns when a local pulled workspace exists and now diverges. Covers the "switch model → run test → switch back" scenario in two commands.
+- **`newo logs --name <ActionName>` (R4)** — client-side filter on `data.name` (e.g. only `Gen` or `GetMemory` action calls), replacing `--json` + jq post-filtering. Help text now documents that the model of a turn must be read from `data.source.model` in `--json` output, not inferred from actor/agent names.
+- **`newo logs --raw`** — JSONL output: one JSON object per log line (sorted oldest-first), with no banners or empty-result text, so stdout is a clean machine-readable stream for piping (`newo logs --raw | jq ...`). Distinct from `--json`, which emits a single pretty-printed JSON array.
+- **`src/sync/remote-skill.ts`** — shared resolver (`resolveRemoteSkill`, `parseModelFlag`) that maps a project/agent/flow/skill IDN path to platform IDs with descriptive errors listing available IDNs at the failing level.
+- **20 unit tests** — 14 in `test/sandbox-skill-commands.test.js` covering connector selection (legacy first-running behavior, exact match, non-running exclusion, error listings), poll timeout override, user/agent `external_event_id` capture, IDN-path resolution, and `--model` flag parsing, plus 6 in `test/logs-command.test.js` covering `--name` client-side filtering, pagination, and `--raw` JSONL output — all against a fake axios client, no network.
+
+## [3.7.4] - 2026-06-10
+
+### Added
+
+- **V2 push now creates new skills from inline flow YAML definitions.** Previously the `newo_v2` push path could only update content of skills already known to the local project map; a skill added locally to `{FlowIdn}.yaml` after the last pull was silently ignored, leaving its `.nsl`/`.nslg` script with no way to reach the platform. `V2ProjectSyncStrategy.push()` now reconciles inline skill metadata before iterating script changes: missing remote skills are created via `POST /api/v1/designer/flows/{flowId}/skills`, race-condition duplicates (already-exists on create) fall back to fetching the remote skill via `listFlowSkills` and updating it, and missing skill parameters are filled in via `POST /api/v1/designer/flows/skills/{skillId}/parameters`. The local project map and SHA256 hash store are updated atomically after the reconciliation pass.
+- **Strict model validation before V2 skill writes.** New `assertSkillModelResolved` helper throws a descriptive error ("Set either skill.model.* or flow default_model_idn/default_provider_idn") when neither the inline skill nor the flow declares a model/provider, instead of letting the platform return a generic 4xx with no hint about which YAML field is missing.
+- **33 unit tests** in `test/v2-push-helpers.test.js` covering `isAlreadyExistsApiError`, `assertSkillModelResolved`, `normalizeRunnerType`, `normalizeParameters`, `skillMetadataDiffers`, `buildV2SkillMetadataFromYaml`, and `createMissingSkillParameters` (stubbed API client). Regression-locks the false-positive matcher fix, the model-validation contract, the parameter-creation counting, and the model-key-order false positive.
+
+### Fixed
+
+- **`isAlreadyExistsApiError` no longer matches "does not exist".** The previous detector accepted any 400/409/422 response whose body contained the substring `"exist"`, which incorrectly classified "Skill does not exist" / "Flow doesn't exist" errors as duplicates and triggered a spurious reuse fallback. The matcher is now tightened to only `"already exists"` and `"duplicate key"`.
+- **`validate()` reports missing scripts for YAML-declared skills.** Previously `V2ProjectSyncStrategy.validate()` only checked scripts listed in the local map; skills added directly to `{FlowIdn}.yaml` since the last pull were never validated. Validation now walks the YAML when present and surfaces a `Script file not found:` error per skill, matching the new push contract.
+- **Missing-script skills are reported, not silently skipped.** Instead of a `verbose`-only "Skipping skill without local script" log, the push now records `[newo_v2] Missing script for skill {project}/{agent}/{flow}/{skill}: {path}` in the push errors. Failures are isolated per skill: one broken skill no longer aborts the push of every other flow/project in the workspace.
+- **Model validation only runs before actual platform writes.** `assertSkillModelResolved` fires only when a skill is about to be created or its metadata updated — a pre-existing flow YAML without `skill.model.*` / flow `default_model_idn` no longer blocks pushes of unrelated, untouched skills.
+- **Parameter-creation count is no longer inflated by already-existing parameters.** `createMissingSkillParameters` incremented its counter (and logged "Created skill parameter") even when the platform answered "already exists", which spuriously triggered a follow-up `updateSkill` call. Count and log now happen only on successful creation.
+- **`isAlreadyExistsApiError` no longer throws on `null`/`undefined` errors** (optional chaining on `.response`), which previously masked the original failure inside catch handlers.
+- **`skillMetadataDiffers` no longer flags every skill as changed.** The model comparison used `JSON.stringify`, which is key-order-sensitive: the project map stores `{provider_idn, model_idn}` (platform API order) while YAML-built metadata uses `{model_idn, provider_idn}`. Verified against a live account: every push rewrote all 1342 skills. Model is now compared field-by-field and parameters are compared order-insensitively (sorted by name).
+- **Skill parameters are now created explicitly after `createSkill`.** The platform's create endpoint ignores inline `parameters` in the request body (verified against the live platform) — newly created skills silently lost their YAML-declared parameters. The create path now calls `POST /flows/skills/{skillId}/parameters` per parameter, same as the update path.
+
 ## [3.7.3] - 2026-05-25
 
 ### Fixed
@@ -1061,7 +1092,9 @@ Another Item: $Price [Modifiers: modifier3]
 - GitHub Actions CI/CD integration
 - Robust authentication with token refresh
 
-[Unreleased]: https://github.com/sabbah13/newo-cli/compare/v3.7.3...HEAD
+[Unreleased]: https://github.com/sabbah13/newo-cli/compare/v3.7.5...HEAD
+[3.7.5]: https://github.com/sabbah13/newo-cli/compare/v3.7.4...v3.7.5
+[3.7.4]: https://github.com/sabbah13/newo-cli/compare/v3.7.3...v3.7.4
 [3.7.3]: https://github.com/sabbah13/newo-cli/compare/v3.7.2...v3.7.3
 [3.7.2]: https://github.com/sabbah13/newo-cli/compare/v3.7.1...v3.7.2
 [3.7.1]: https://github.com/sabbah13/newo-cli/compare/v3.7.0...v3.7.1
@@ -1095,4 +1128,4 @@ Another Item: $Price [Modifiers: modifier3]
 [1.4.0]: https://github.com/sabbah13/newo-cli/compare/v1.2.1...v1.4.0
 [1.3.0]: https://github.com/sabbah13/newo-cli/compare/v1.2.1...v1.3.0
 [1.2.2]: https://github.com/sabbah13/newo-cli/compare/v1.2.1...v1.2.2
-[1.2.1]: https://github.com/sabbah13/newo-cli/releases/tag/v1.2.1
+[1.2.1]: https://github.com/sabbah13/newo-cli/releases/tag/v1.2.1

package/README.md

@@ -8,6 +8,9 @@
 **NEWO CLI** - Professional command-line tool for NEWO AI Agent development. Features **modular architecture**, **IDN-based file management**, and **comprehensive multi-customer support**.
 
 Sync NEWO "Project → Agent → Flow → Skills" structure to local files with:
+- 🆕 **V2 skill creation on push** (v3.7.4) - adding a skill inline to a `newo_v2` `{FlowIdn}.yaml` and pushing now creates it on the platform (previously only updates of existing skills worked)
+- 🆕 **Sandbox connector selection + automation** (v3.7.5) - `newo sandbox --connector <idn>`, `--list-connectors`, `--file`/`--stdin` for long messages, `--timeout`, and `--json` with `external_event_id` for log correlation
+- 🆕 **Point skill edits** (v3.7.5) - `newo get-skill` / `newo update-skill` inspect and modify a single skill (model, script) on the platform without a pulled workspace, with optional `--publish`
 - 🆕 **Canvas blank-screen hardening** (v3.7.3) - JSON-typed attributes (e.g. Workflow Builder canvas) with Markdown `\_` escapes or structural newlines no longer corrupt the canvas on push ([#7](https://github.com/sabbah13/newo-cli/pull/7))
 - 🆕 **Flow metadata sync** (v3.7.2) - `newo push` now reconciles flow title, events, and state_fields from local `metadata.yaml` to the platform (closes [#3](https://github.com/sabbah13/newo-cli/issues/3))
 - 🆕 **Dual format support** (v3.6.0) - `cli_v1` (native) and `newo_v2` (platform compatible), auto-detected per customer
@@ -149,12 +152,78 @@ NEWO_REFRESH_URL=custom_refresh_endpoint   # Custom refresh endpoint
 | `newo push [--format <fmt>]` | Upload local changes to NEWO | • Works with both formats<br>• Hash-based change detection<br>• Library skill updates<br>• Publishes flows automatically |
 | `newo status [--format <fmt>]` | Show modified files | • Format-aware status<br>• Multiple file warnings<br>• Per-customer status |
 | `newo export [--output <file>]` | Download V2 bulk ZIP from platform | • Complete organization export<br>• Projects, agents, flows, skills, attributes, AKB<br>• Compatible with platform UI import |
-| `newo sandbox` | Test agents in sandbox chat mode | • Single-command mode for automation<br>• Multi-turn conversation support<br>• Debug info for agent development |
+| `newo sandbox` | Test agents in sandbox chat mode | • Single-command mode for automation<br>• Multi-turn conversation support<br>• Debug info for agent development<br>• v3.7.5: `--connector`, `--list-connectors`, `--file`/`--stdin`, `--timeout`, `--json` |
+| `newo get-skill` / `newo update-skill` | Inspect / point-edit one skill on the platform (NEW v3.7.5) | • No pulled workspace required<br>• `--model <provider>/<model>`, `--script <file>`<br>• Optional `--publish` |
 | `newo conversations` | Pull conversation history | • User personas and chat history<br>• YAML format output<br>• Pagination support |
 | `newo list-customers` | List configured customers | • Shows default customer<br>• Multi-customer discovery |
 | `newo import-akb` | Import knowledge base articles | • Structured text parsing<br>• Bulk article import<br>• Validation and error reporting |
 | `newo meta` | Get project metadata (debug) | • Project structure analysis<br>• Metadata validation |
 
+### Sandbox Connector Selection & Automation (NEW v3.7.5)
+
+`newo sandbox` previously always chatted through the **first** running connector of the `sandbox` integration, making other agents (e.g. a Vibe Builder behind a `vibe_agent` connector) unreachable from the CLI. v3.7.5 adds connector selection plus automation-friendly I/O:
+
+```bash
+newo sandbox --list-connectors                          # show running sandbox connectors
+newo sandbox "ping" --connector vibe_agent              # chat through a specific connector
+newo sandbox "ping" --integration sandbox --connector vibe_agent   # explicit pair (default integration: sandbox)
+
+newo sandbox --file ./chunk1.txt --actor <id> --json    # long message from file, machine-readable output
+cat chunk1.txt | newo sandbox --stdin --actor <id>      # message from stdin
+newo sandbox "ping" --timeout 420                       # wait up to 7 minutes for slow agent turns (default: 60s)
+```
+
+`--json` output includes the correlation keys for `newo logs`:
+
+```json
+{
+  "actor_id": "…", "persona_id": "…", "connector_idn": "vibe_agent",
+  "external_event_id": "…",          // user turn — use with: newo logs --event-id <id>
+  "user_external_event_id": "…", "agent_external_event_id": "…",
+  "response": "…", "elapsed_ms": 12345, "timed_out": false,
+  "flow_idn": "VibeFlow", "skill_idn": "…", "session_id": "…"
+}
+```
+
+Without `--connector`, behavior is unchanged (first running connector), so existing scripts keep working. When the requested connector is missing, the error lists the available running connectors.
+
+This makes agent integration tests scriptable in plain bash: send chunks via `newo sandbox --file … --json`, then verify behavior via `newo logs --event-id … --json` — no more ad-hoc Node scripts importing `dist/api.js`.
+
+### Point Skill Inspection & Edits (NEW v3.7.5)
+
+Inspect or modify a **single skill** directly on the platform by IDN path — no pulled `newo_customers/` workspace required, nothing else gets touched:
+
+```bash
+# What lives on the platform right now? (model, runner_type, parameters, prompt_script)
+newo get-skill structured_generation --project vibe --agent VibeAgent --flow VibeFlow [--json]
+
+# Temporarily switch a model, publish, run tests, switch back:
+newo update-skill structured_generation --project vibe --agent VibeAgent --flow VibeFlow \
+    --model google/gemini25_pro --publish
+# …run your tests…
+newo update-skill structured_generation --project vibe --agent VibeAgent --flow VibeFlow \
+    --model openai/gpt54 --publish
+
+# Replace a skill script from a file (draft only, without --publish):
+newo update-skill get_memory --project vibe --agent VibeAgent --flow VibeFlow --script ./patched.nsl
+```
+
+- `--model <provider_idn>/<model_idn>` and `--script <file>` can be combined; everything else (title, parameters, runner) is preserved.
+- `--publish` publishes the flow after the update (same as push); without it the change stays draft. `--publish-description "<text>"` sets the publish note.
+- If a pulled local workspace exists for the project, the CLI warns that it now diverges from the platform.
+
+### Logs: Action-Name Filter (NEW v3.7.5)
+
+```bash
+newo logs --type call --name Gen --json        # only Gen action calls
+newo logs --name GetMemory --hours 24          # only GetMemory calls, last 24h
+newo logs --type call --raw | jq '.data.name'  # JSONL stream, one log per line, for piping
+```
+
+`--name` filters by `data.name` client-side (the API has no such query param). Note: the **model used for a turn** is in `data.source.model` of the `--json` output — do not infer it from actor/agent names.
+
+`--raw` emits **JSONL** — one JSON object per log line (oldest-first), with no banners or "no logs found" text — so stdout is a clean stream for `jq`/piping. This differs from `--json`, which prints a single pretty-printed JSON array.
+
 ### Flow Metadata Sync (NEW v3.7.2)
 
 `newo push` now reconciles **flow-level metadata** — title, `events:`, and `state_fields:` — from local YAML to the platform. Before v3.7.2 push only uploaded skill scripts, so edits to a flow's `metadata.yaml` (V1) or `{FlowIdn}.yaml` (V2) silently never reached the platform; events added via `newo create-event` could appear to disappear after a pull → push cycle. Closes [#3](https://github.com/sabbah13/newo-cli/issues/3).
@@ -1378,4 +1447,4 @@ NEWO CLI integrates with these NEWO platform endpoints:
 
 ---
 
-**Built with ❤️ by the NEWO team**
+**Built with ❤️ by the NEWO team**

package/dist/cli/commands/get-skill.d.ts

@@ -0,0 +1,3 @@
+import type { MultiCustomerConfig, CliArgs } from '../../types.js';
+export declare function handleGetSkillCommand(customerConfig: MultiCustomerConfig, args: CliArgs, verbose?: boolean): Promise<void>;
+//# sourceMappingURL=get-skill.d.ts.map

package/dist/cli/commands/get-skill.js

@@ -0,0 +1,72 @@
+/**
+ * Get Skill Command Handler - Inspect a skill's live state on the platform
+ *
+ * Usage:
+ *   newo get-skill <skill-idn> --project <idn> --agent <idn> --flow <idn> [--json]
+ *
+ * Shows the skill as it currently lives on the platform (model, runner_type,
+ * parameters, prompt_script) without requiring a pulled local workspace.
+ */
+import { requireSingleCustomer } from '../customer-selection.js';
+import { makeClient } from '../../api.js';
+import { getValidAccessToken } from '../../auth.js';
+import { resolveRemoteSkill } from '../../sync/remote-skill.js';
+const USAGE = 'Usage: newo get-skill <skill-idn> --project <project-idn> --agent <agent-idn> --flow <flow-idn> [--json] [--customer <idn>]';
+export async function handleGetSkillCommand(customerConfig, args, verbose = false) {
+    const skillIdn = args._[1];
+    const projectIdn = args.project;
+    const agentIdn = args.agent;
+    const flowIdn = args.flow;
+    const asJson = Boolean(args.json);
+    if (!skillIdn || !projectIdn || !agentIdn || !flowIdn) {
+        console.error('Error: skill IDN, --project, --agent and --flow are required');
+        console.error(USAGE);
+        process.exit(1);
+    }
+    const selectedCustomer = requireSingleCustomer(customerConfig, args.customer);
+    if (asJson) {
+        process.env.NEWO_QUIET_MODE = 'true'; // keep stdout machine-readable for piping
+    }
+    const token = await getValidAccessToken(selectedCustomer);
+    const client = await makeClient(verbose, token);
+    if (verbose)
+        console.log(`🔍 Resolving skill ${projectIdn}/${agentIdn}/${flowIdn}/${skillIdn}...`);
+    const { project, agent, flow, skill } = await resolveRemoteSkill(client, {
+        projectIdn,
+        agentIdn,
+        flowIdn,
+        skillIdn
+    });
+    if (asJson) {
+        console.log(JSON.stringify({
+            project_idn: project.idn,
+            agent_idn: agent.idn,
+            flow_idn: flow.idn,
+            id: skill.id,
+            idn: skill.idn,
+            title: skill.title,
+            runner_type: skill.runner_type,
+            model: skill.model,
+            parameters: skill.parameters,
+            prompt_script: skill.prompt_script ?? null
+        }, null, 2));
+        return;
+    }
+    console.log(`📜 Skill: ${project.idn}/${agent.idn}/${flow.idn}/${skill.idn}`);
+    console.log(`   ID: ${skill.id}`);
+    console.log(`   Title: ${skill.title}`);
+    console.log(`   Runner type: ${skill.runner_type}`);
+    console.log(`   Model: ${skill.model.provider_idn}/${skill.model.model_idn}`);
+    if (skill.parameters.length > 0) {
+        console.log(`   Parameters:`);
+        for (const param of skill.parameters) {
+            console.log(`     ${param.name}${param.default_value !== undefined ? ` = ${JSON.stringify(param.default_value)}` : ''}`);
+        }
+    }
+    else {
+        console.log(`   Parameters: (none)`);
+    }
+    console.log(`\n--- prompt_script (${(skill.prompt_script || '').length} chars) ---`);
+    console.log(skill.prompt_script || '(empty)');
+}
+//# sourceMappingURL=get-skill.js.map

package/dist/cli/commands/help.js

@@ -16,6 +16,11 @@ Core Commands:
   newo conversations [--customer <idn>] [--all] # download user conversations -> conversations.yaml
   newo sandbox "<message>" [--customer <idn>]   # test agent in sandbox - single message mode
   newo sandbox --actor <id> "message"           # continue existing sandbox conversation
+  newo sandbox --list-connectors                # list running sandbox connectors (NEW v3.7.5)
+  newo sandbox "<msg>" --connector <idn>        # chat through a specific connector (NEW v3.7.5)
+  newo sandbox --file <path> | --stdin          # send long messages from file/stdin (NEW v3.7.5)
+  newo get-skill <idn> --project <p> --agent <a> --flow <f> [--json]   # inspect live skill on platform (NEW v3.7.5)
+  newo update-skill <idn> --project <p> --agent <a> --flow <f> [--model <prov>/<model>] [--script <file>] [--publish]  # point-edit skill (NEW v3.7.5)
   newo pull-attributes [--customer <idn>]       # download customer + project attributes
   newo list-customers                           # list available customers
   newo meta [--customer <idn>]                  # get project metadata (debug)
@@ -76,6 +81,8 @@ Analytics & Monitoring (NEW):
   newo logs --type <types>                                   # filter by type: system, operation, call (comma-separated)
   newo logs --flow <idn> --skill <idn>                       # filter by flow and/or skill
   newo logs --message <text>                                 # search in log messages
+  newo logs --name <ActionName>                              # filter by action name in data.name, e.g. Gen, GetMemory (NEW v3.7.5)
+  newo logs --event-id <uuid>                                # filter by external event ID (correlate with sandbox --json)
   newo logs --follow, -f                                     # tail mode - continuously poll for new logs
   newo logs --json                                           # output logs as JSON array
   newo logs --raw                                            # output each log as single JSON line (for piping)
@@ -94,6 +101,12 @@ Flags:
   --verbose, -v                # enable detailed logging and progress information
   --quiet, -q                  # minimal output for automation (sandbox only)
   --actor <id>                 # continue existing sandbox chat with actor/chat ID
+  --connector <idn>            # sandbox: select connector by connector_idn (default: first running)
+  --integration <idn>          # sandbox: integration to search connectors in (default: sandbox)
+  --file <path>                # sandbox: read message text from file
+  --stdin                      # sandbox: read message text from stdin
+  --timeout <seconds>          # sandbox: max wait for agent response (default: 60)
+  --json                       # sandbox: machine-readable output incl. external_event_id (user+agent turns)
   --confirm                    # confirm destructive operations without prompting
   --no-publish                 # skip automatic flow publishing during push operations
   --output, -o <file>          # output file path (for export command)
@@ -201,6 +214,22 @@ Usage Examples:
   newo sandbox "Test query" --verbose                            # With debug info
   newo sandbox "Test query" --quiet                              # For automation/scripts
 
+  # Sandbox connector selection + automation (NEW v3.7.5):
+  newo sandbox --list-connectors                                 # Show running sandbox connectors
+  newo sandbox "ping" --connector vibe_agent                     # Chat through specific connector
+  newo sandbox --file ./chunk1.txt --actor abc123... --json      # Long message from file, JSON output
+  cat msg.txt | newo sandbox --stdin --timeout 420 --json        # From stdin with 7-minute timeout
+  # --json output: {actor_id, persona_id, external_event_id, user_external_event_id,
+  #                 agent_external_event_id, response, elapsed_ms, timed_out, ...}
+  # Correlate a turn with its logs: newo logs --event-id <external_event_id> --json
+
+  # Live skill inspection / point edits (NEW v3.7.5):
+  newo get-skill structured_generation --project vibe --agent VibeAgent --flow VibeFlow            # View live state
+  newo get-skill structured_generation --project vibe --agent VibeAgent --flow VibeFlow --json     # As JSON
+  newo update-skill structured_generation --project vibe --agent VibeAgent --flow VibeFlow \\
+      --model openai/gpt54 --publish                             # Switch model and publish
+  newo update-skill my_skill --project p --agent a --flow f --script ./patched.nsl --publish       # Replace script
+
   # Analytics logs (NEW v3.5.0):
   newo logs                                                      # Last 1 hour of logs
   newo logs --hours 24                                           # Last 24 hours

package/dist/cli/commands/logs.d.ts

@@ -12,7 +12,12 @@
  *   newo logs --follow                     # Tail mode (poll for new logs)
  *   newo logs --json                       # Output as JSON
  */
-import type { MultiCustomerConfig, CliArgs } from '../../types.js';
+import type { AxiosInstance } from 'axios';
+import type { MultiCustomerConfig, LogEntry, LogsQueryParams, LogsResponse, CliArgs } from '../../types.js';
+type GetLogsFn = (client: AxiosInstance, params: LogsQueryParams) => Promise<LogsResponse>;
 export declare function handleLogsCommand(customerConfig: MultiCustomerConfig, args: CliArgs, verbose: boolean): Promise<void>;
+export declare function collectLogsForDisplay(client: AxiosInstance, params: LogsQueryParams, nameFilter?: string | null, getLogsFn?: GetLogsFn): Promise<LogEntry[]>;
+export declare function fetchAndDisplayLogs(client: AxiosInstance, params: LogsQueryParams, asJson: boolean, raw: boolean, nameFilter?: string | null, getLogsFn?: GetLogsFn): Promise<void>;
 export declare function printLogsHelp(): void;
+export {};
 //# sourceMappingURL=logs.d.ts.map

package/dist/cli/commands/logs.js

@@ -85,7 +85,14 @@ function formatLogEntryCompact(log, showColors = true) {
 export async function handleLogsCommand(customerConfig, args, verbose) {
     // Select customer
     const selectedCustomer = requireSingleCustomer(customerConfig, args.customer);
-    console.log(`📊 Fetching logs for ${selectedCustomer.idn}...`);
+    // Keep stdout machine-readable when JSON output is requested (for piping to jq etc.)
+    const machineOutput = Boolean(args.json || args.raw);
+    if (machineOutput) {
+        process.env.NEWO_QUIET_MODE = 'true'; // suppress auth logging on stdout
+    }
+    else {
+        console.log(`📊 Fetching logs for ${selectedCustomer.idn}...`);
+    }
     // Get access token and create client
     const token = await getValidAccessToken(selectedCustomer);
     const client = await makeClient(verbose, token);
@@ -146,21 +153,59 @@ export async function handleLogsCommand(customerConfig, args, verbose) {
     const follow = Boolean(args.follow || args.f);
     const asJson = Boolean(args.json);
     const raw = Boolean(args.raw);
+    // --name filters by data.name (e.g. action name like Gen or GetMemory).
+    // The API has no such query param, so it is applied client-side.
+    const nameFilter = args.name ? String(args.name) : null;
     if (follow) {
-        await tailLogs(client, params, asJson);
+        await tailLogs(client, params, asJson, nameFilter);
     }
     else {
-        await fetchAndDisplayLogs(client, params, asJson, raw);
+        await fetchAndDisplayLogs(client, params, asJson, raw, nameFilter);
+    }
+}
+function filterByName(logs, nameFilter) {
+    if (!nameFilter)
+        return [...logs];
+    return logs.filter(log => log.data['name'] === nameFilter);
+}
+export async function collectLogsForDisplay(client, params, nameFilter = null, getLogsFn = getLogs) {
+    if (!nameFilter) {
+        const response = await getLogsFn(client, params);
+        return [...response.items];
+    }
+    const pageSize = Number.isFinite(params.per) && params.per && params.per > 0 ? params.per : 50;
+    let page = Number.isFinite(params.page) && params.page && params.page > 0 ? params.page : 1;
+    const logs = [];
+    while (true) {
+        const response = await getLogsFn(client, {
+            ...params,
+            page,
+            per: pageSize
+        });
+        logs.push(...filterByName(response.items, nameFilter));
+        if (response.items.length < pageSize) {
+            break;
+        }
+        page++;
     }
+    return logs;
 }
-async function fetchAndDisplayLogs(client, params, asJson, raw) {
+export async function fetchAndDisplayLogs(client, params, asJson, raw, nameFilter = null, getLogsFn = getLogs) {
     try {
-        const response = await getLogs(client, params);
-        const logs = response.items;
+        const logs = await collectLogsForDisplay(client, params, nameFilter, getLogsFn);
         if (asJson) {
             console.log(JSON.stringify(logs, null, 2));
             return;
         }
+        // --raw is a machine-readable JSONL contract: stdout must contain only
+        // one JSON object per log line, with no banners or empty-result text.
+        if (raw) {
+            const sortedLogs = [...logs].sort((a, b) => new Date(a.datetime).getTime() - new Date(b.datetime).getTime());
+            for (const log of sortedLogs) {
+                console.log(JSON.stringify(log));
+            }
+            return;
+        }
         if (logs.length === 0) {
             console.log('\nNo logs found for the specified criteria.');
             return;
@@ -171,12 +216,7 @@ async function fetchAndDisplayLogs(client, params, asJson, raw) {
         // Detect if stdout is a TTY (supports colors)
         const useColors = process.stdout.isTTY !== false;
         for (const log of sortedLogs) {
-            if (raw) {
-                console.log(JSON.stringify(log));
-            }
-            else {
-                console.log(formatLogEntry(log, useColors));
-            }
+            console.log(formatLogEntry(log, useColors));
         }
         console.log(`\n✅ Displayed ${logs.length} log entries`);
     }
@@ -185,7 +225,7 @@ async function fetchAndDisplayLogs(client, params, asJson, raw) {
         console.error('Failed to fetch logs:', err.response?.status, err.response?.data || err.message);
     }
 }
-async function tailLogs(client, params, asJson) {
+async function tailLogs(client, params, asJson, nameFilter = null) {
     console.log('🔄 Watching for new logs (Ctrl+C to stop)...\n');
     const seenLogIds = new Set();
     let lastCheckTime = params.from_datetime || new Date(Date.now() - 60 * 60 * 1000).toISOString();
@@ -201,7 +241,7 @@ async function tailLogs(client, params, asJson) {
                 per: 100
             };
             const response = await getLogs(client, pollParams);
-            const logs = response.items;
+            const logs = filterByName(response.items, nameFilter);
             // Filter out already seen logs and sort by time
             const newLogs = logs
                 .filter(log => !seenLogIds.has(log.log_id))
@@ -257,6 +297,7 @@ Filter Options:
   --flow <idn>          Filter by flow IDN
   --skill <idn>         Filter by skill IDN
   --message <text>      Search in log messages
+  --name <ActionName>   Filter by action name in data.name, e.g. Gen, GetMemory (client-side)
   --event-id <uuid>     Filter by external event ID
   --runtime-id <uuid>   Filter by runtime context ID
   --actor-id <uuid>     Filter by user actor ID
@@ -278,6 +319,13 @@ Examples:
   newo logs --type call --skill CreateActor    # Skill calls for CreateActor
   newo logs --flow CACreatorFlow --follow      # Tail logs for specific flow
   newo logs --json --per 100                   # Get 100 logs as JSON
+  newo logs --type call --name Gen --json      # Only Gen action calls
+
+Notes:
+  The model used for a turn is in data.source.model of the --json output —
+  do NOT infer it from actor/agent names.
+  external_event_id (newo sandbox --json) correlates a chat turn with its
+  logs: newo logs --event-id <id>
 `);
 }
 //# sourceMappingURL=logs.js.map

package/dist/cli/commands/sandbox.d.ts

@@ -1,14 +1,20 @@
 /**
  * Sandbox Chat Command Handler
  * Supports both single-command and interactive modes
+ *
+ * Usage:
+ *   npx newo sandbox "Hello" --customer <idn>               # Single message mode
+ *   npx newo sandbox --actor <actor_id> "Follow up"         # Continue existing chat
+ *   npx newo sandbox "ping" --connector vibe_agent          # Select specific connector (v3.7.5)
+ *   npx newo sandbox --list-connectors                      # Show running sandbox connectors (v3.7.5)
+ *   npx newo sandbox --file ./msg.txt --actor <id>          # Message from file (v3.7.5)
+ *   cat msg.txt | npx newo sandbox --stdin                  # Message from stdin (v3.7.5)
+ *   npx newo sandbox "ping" --timeout 420                   # Custom response timeout in seconds (v3.7.5)
+ *   npx newo sandbox "ping" --json                          # Machine-readable output (v3.7.5)
  */
 import type { MultiCustomerConfig, CliArgs } from '../../types.js';
 /**
  * Handle sandbox command
- * Usage:
- *   npx newo sandbox "Hello" --customer <idn>              # Single message mode
- *   npx newo sandbox --actor <actor_id> "Follow up"       # Continue existing chat
- *   npx newo sandbox --interactive                        # Interactive mode (TBD)
  */
 export declare function handleSandboxCommand(customerConfig: MultiCustomerConfig, args: CliArgs, verbose: boolean): Promise<void>;
 //# sourceMappingURL=sandbox.d.ts.map