@octavus/agent 1.0.11 → 1.0.12
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +7 -0
- package/dist/index.js +46 -42
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -47,6 +47,7 @@ The working directory you launch in is the agent's workspace (override with `--w
|
|
|
47
47
|
- `--backup-model <provider/model-id>` - backup model for this run only.
|
|
48
48
|
- `--capability <slug>=<on|off>` - toggle one capability for this run (repeatable).
|
|
49
49
|
- `--config <file>` - JSON run config for scripted sweeps (see below).
|
|
50
|
+
- `--benchmark-tools` - enable the opt-in benchmark tools (e.g. `ask_human`) for this run (see below).
|
|
50
51
|
- `--json` - print one machine-readable JSON result to stdout.
|
|
51
52
|
- `--force` - allow running with the workspace at `$HOME` or `/`.
|
|
52
53
|
- `--verbose` - print diagnostics to stderr.
|
|
@@ -87,6 +88,12 @@ Precedence: explicit flags > `--config` file > the agent's dashboard defaults. A
|
|
|
87
88
|
|
|
88
89
|
`--json` prints exactly one JSON object to stdout - `{ "threadId", "sessionId", "status", "threadUrl" }` - with all human/progress output routed to stderr, so a benchmark harness can capture and correlate a run without scraping text. The exit code signals completed (`0`), failed (`1`), or stopped from the dashboard (`4`). Read the full transcript and a per-run cost/usage summary from `GET /api/v1/workforce/agents/{agentId}/threads/{threadId}` with the same agent key.
|
|
89
90
|
|
|
91
|
+
### Benchmark tools
|
|
92
|
+
|
|
93
|
+
`--benchmark-tools` provides an extra, opt-in set of tools for benchmark harnesses that need a capability the agent's normal toolset lacks. It is off by default and never present on a normal run, so it changes nothing unless you pass it.
|
|
94
|
+
|
|
95
|
+
Today it adds one tool, `benchmark__ask_human`, which lets the agent ask the operator driving the run for an input only they can provide (a withheld form field, a business rule, the location of a source file) and receive the answer inline: the call blocks until the harness answers, then returns the answer as the tool result and the run continues. The agent runs inside the machine you control, so the harness answers by reading `benchmark-ask.json` and writing `benchmark-answer.json` in the run's workdir - an internal transport the agent never sees. Enable it only when your harness is set up to answer; otherwise the tool waits briefly and then tells the agent to proceed on its own.
|
|
96
|
+
|
|
90
97
|
### Environments
|
|
91
98
|
|
|
92
99
|
`--env <name>` (or `OCTAVUS_ENV`, default `production`) selects an environment. Each keeps its own credentials and browser profile under `~/.octavus/<env>/`, so you can configure and run more than one side by side without them clobbering each other:
|