tasqmate 0.2.0 → 0.3.1
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 +18 -7
- package/dist/tasqmate.js +1715 -555
- package/package.json +2 -2
package/README.md
CHANGED
|
@@ -49,7 +49,8 @@ tasqmate doc archive <doc> [--restore]
|
|
|
49
49
|
tasqmate comment add <ref> <body> [--reply-to <commentRef>] [--mention <name>]…
|
|
50
50
|
tasqmate comment add --doc <docId> <body> [--thread <threadId>] [--propose <text>]
|
|
51
51
|
tasqmate comment add <ref> <body> --attach <path>…
|
|
52
|
-
tasqmate file add <ref> <path>… | list <ref> | remove <id>
|
|
52
|
+
tasqmate file add <ref> <path>… [--inline] | list <ref> | remove <id>
|
|
53
|
+
tasqmate file add --doc <doc> <path>… [--inline] # --inline prints the Markdown line that shows it in the body
|
|
53
54
|
tasqmate file get <id> [--out <path>] [--force] [--link]
|
|
54
55
|
tasqmate time log <ref> <duration> [--day YYYY-MM-DD] [--note <text>] [--billable yes|no]
|
|
55
56
|
tasqmate time list <ref> | edit <entryId> [--duration] [--day] [--note | --no-note] [--billable] | delete <entryId>
|
|
@@ -119,23 +120,32 @@ Both are examples. Tasqmate stores and delivers the event; it never runs your ag
|
|
|
119
120
|
|
|
120
121
|
## Keep an agent running on your machine (`tasqmate bridge`)
|
|
121
122
|
|
|
122
|
-
The bridge runs Claude or
|
|
123
|
+
The bridge runs Claude, Codex or Hermes on your own computer as one Tasqmate agent, on your own sign-in. It long-polls the agent's wakes (mentions, replies, assignments), starts your harness in the project's folder for each one, and the agent answers in Tasqmate through the Tasqmate MCP server. Tasqmate never runs a model and never sees your Claude, ChatGPT or model-provider credentials.
|
|
123
124
|
|
|
124
125
|
```sh
|
|
125
|
-
claude auth login # or: codex login (you sign in; the bridge never does)
|
|
126
|
+
claude auth login # or: codex login, or hermes setup (you sign in; the bridge never does)
|
|
126
127
|
tasqmate bridge setup --map "Homepage=~/code/homepage"
|
|
127
128
|
```
|
|
128
129
|
|
|
129
|
-
`setup` finds `claude` and `
|
|
130
|
+
`setup` finds `claude`, `codex` and `hermes`, checks their sign-in with `claude auth status` / `codex login status` / Hermes's ACP handshake only, prints a code to approve on the pair page (an admin or the agent's owner picks the agent), keeps the agent key in the OS keychain (macOS Keychain, libsecret, Windows Credential Manager; a 0600 file when there is none), saves the pairing, maps project folders (`~` is your home folder; a folder that isn't there is asked for again, and stopping here keeps the pairing so `bridge map` can finish) and installs a background service (launchd, a systemd user unit, or a logon task). Claude runs through the Claude Agent SDK, which `setup` installs for the bridge alone on first use.
|
|
131
|
+
|
|
132
|
+
Hermes runs as `hermes acp` (the Agent Client Protocol) with your own Hermes setup: its tools are the ones your Hermes config gives ACP. The bridge answers what Hermes asks about: a file edit is allowed only inside the mapped folder, and a command Hermes flags as needing approval is refused, because nobody is there to approve it. What Hermes does not ask about (reading files, ordinary commands, web tools you enabled) runs as it would in your editor, so pick that setup with care, or give Hermes a profile with fewer tools.
|
|
130
133
|
|
|
131
134
|
| Command | What |
|
|
132
135
|
| --- | --- |
|
|
133
|
-
| `bridge setup` | Pair, map folders, install the service. `--harness claude\|codex`, `--map project=folder` (repeat), `--no-service`, `--runs-per-hour`, `--timeout`, `--max-turns` |
|
|
136
|
+
| `bridge setup` | Pair, map folders, install the service. `--harness claude\|codex\|hermes`, `--map project=folder` (repeat), `--no-service`, `--runs-per-hour`, `--timeout`, `--max-turns`, `--replace` (pair again without asking when this machine is already paired), `--resume` (finish a setup that stopped after pairing) |
|
|
134
137
|
| `bridge start` | Run in the foreground (what the service runs) |
|
|
135
|
-
| `bridge status` | Agent, harness sign-in, server presence, service, folders, limits |
|
|
138
|
+
| `bridge status` | Agent, harness sign-in, server presence, service, folders, limits, and a Next step line when something is off |
|
|
139
|
+
| `bridge doctor [--fix]` | Check everything in order (CLI, pairing, key, Tasqmate, harness, service, folders) and print the fix; `--fix` applies the safe ones after asking; exits 1 when something is wrong |
|
|
136
140
|
| `bridge logs [-n 50] [-f]` | The bridge's log |
|
|
137
141
|
| `bridge map <project> [folder]` | Map a project to a folder (`--remove` to drop it); read on the next wake |
|
|
138
|
-
| `bridge unpair` | Revoke the key on the server, remove the service, key and config |
|
|
142
|
+
| `bridge unpair` | Revoke the key on the server, remove the service, key and config; on an unpaired machine, revoke and delete a key an unfinished setup left |
|
|
143
|
+
|
|
144
|
+
### When setup goes wrong
|
|
145
|
+
|
|
146
|
+
Running `tasqmate bridge setup` again always has a way forward. If the agent was unpaired in Tasqmate, setup clears this machine and pairs again. If this machine is still paired and live, it asks before replacing the pairing (`--replace` skips the question). If an earlier setup stopped after you approved the code, it finds the key that setup left and finishes it here with no new code (`--resume`). Not sure what is wrong? `tasqmate bridge doctor` checks each part and prints the one command that fixes it. Moving an agent to another machine works from the other side too: run setup on the new machine, pick the agent on the pair page and choose Replace.
|
|
147
|
+
|
|
148
|
+
One bridge runs one agent. To run a second agent on the same machine (say Hermes beside Claude), give it a profile: `tasqmate bridge --profile hermes setup --harness hermes`. A profile keeps its own config in `~/.config/tasqmate/bridge-hermes/` and its own background service, and every other bridge command takes the same `--profile hermes` (`bridge --profile hermes status`, `doctor`, `logs`, `map`, `unpair`). Without `--profile` the commands act on the default bridge, as before.
|
|
139
149
|
|
|
140
150
|
What a run may do: read and edit files inside the mapped folder, and use its Tasqmate tools. Shell commands, web access, files outside the folder, and `.git/`, `.claude/` and `.codex/` inside it are refused. A wake on a project with no folder runs in an empty scratch folder, so the agent can still answer but cannot touch code. One run at a time, at most `--runs-per-hour` (default 6) in any hour, each stopped after `--timeout` minutes (30) and `--max-turns` (40). A wake is acknowledged only after its run succeeds; a failed or interrupted run is offered again ten minutes later, up to three times.
|
|
141
151
|
|
|
@@ -148,6 +158,7 @@ What a run may do: read and edit files inside the mapped folder, and use its Tas
|
|
|
148
158
|
| `TASQMATE_WORKSPACE` | Default workspace slug |
|
|
149
159
|
| `TASQMATE_API_URL` | API origin; defaults to `https://tasqmate.pxlabz.com` |
|
|
150
160
|
| `TASQMATE_CONFIG_DIR` | Where the stored token lives; defaults to `$XDG_CONFIG_HOME/tasqmate` |
|
|
161
|
+
| `TASQMATE_BRIDGE_PROFILE` | The bridge profile, same as `tasqmate bridge --profile <name>` |
|
|
151
162
|
|
|
152
163
|
## Exit codes
|
|
153
164
|
|