Add one-click Oddish task runs

README.md

@@ -59,6 +59,9 @@ Trials can run on any provider:
 export OPENAI_API_KEY=<api-key>          # codex CLI + any openai/* model
 export GEMINI_API_KEY=<api-key>          # gemini CLI + any gemini/* model
 export ZAI_API_KEY=<api-key>             # GLM models via the mini-SWE-agent harness
+
+# Optional: one-click hosted trial + public experiment link from the dashboard
+export ODDISH_API_KEY=<full-scope-key>
 ```
 
 ## Usage
@@ -84,6 +87,7 @@ calibration but are sound are exported to `out/easy/`.
 
 Re-running the same command resumes the run from wherever it parked. Ctrl-C stops active model
 trials, keeps completed trials, and leaves unfinished work ready to resume.
+The local dashboard starts automatically and opens the task page as soon as the run is created.
 
 <details>
 <summary>Options</summary>
@@ -98,7 +102,9 @@ trials, keeps completed trials, and leaves unfinished work ready to resume.
 - `--config FILE.json` / `--preset NAME` — Full RunConfig (agents + per-model bands)
 - `--brief TEXT` — Steer the task generation scope (eg "port the FFT subsystem...")
 - `--review` — Pause at the two human gates (scope pick, final QA) instead of auto
+- `--draft` — Export after Static CI with no sweeps or calibration
 - `--yes` — Skip the cost preview confirmation
+- `--no-open-dashboard` — Do not open the task page in a browser
 - `--runs-dir PATH` — Choose the directory for runs (default: `.programsmith/runs`)
 - `--allow-copyleft` — Allow copyleft-licensed sources
 
@@ -133,9 +139,12 @@ programsmith serve
 programsmith stop
 ```
 
-Serves the local dashboard at `http://localhost:8765`: live pipeline DAG for each run, agent
-output, sweep results, file explorer, and optional review gates. Evaluation sweeps remain parked
-unless `serve --spend` is used.
+Serves the local dashboard at `http://localhost:8765`. Each exported task has a direct download
+button and can launch one low-priority Oddish trial, show its agent trajectory, and return a public
+experiment link. Build diagnostics, files, sweep results, and optional review gates remain
+available without dominating the task page. Add a full-scope Oddish API key in Settings to enable
+the hosted run; Oddish free-plan limits apply. Evaluation sweeps remain parked unless
+`serve --spend` is used.
 `programsmith serve` returns after the dashboard is healthy; `programsmith stop` is the explicit
 shutdown command.
 

pyproject.toml

@@ -28,6 +28,7 @@ dependencies = [
     "rich>=13",
     "fastapi>=0.110",
     "uvicorn>=0.29",
+    "httpx>=0.27",
 ]
 
 [project.urls]

src/programsmith/cli.py

@@ -702,6 +702,17 @@ def _ensure_dashboard(runs_dir: Path, host: str, port: int, *, run_key: str | No
     return f"{base}/run/{run_key}" if run_key else base
 
 
+def _open_dashboard(url: str) -> bool:
+    """Open a local dashboard without ever making task creation depend on desktop integration."""
+    if os.getenv("CI") or os.getenv("PROGRAMSMITH_NO_BROWSER"):
+        return False
+    try:
+        import webbrowser
+        return bool(webbrowser.open(url, new=2))
+    except Exception:  # noqa: BLE001 — missing browser/headless sessions are expected
+        return False
+
+
 def _cmd_purge_synthetic(args: argparse.Namespace) -> int:
     """Strip ALL synthetic data from a run and roll the FSM back to its last REAL stage. Removes any
     manifest oracle/sweep entry tagged `simulated`, deletes a synthetic-built task skeleton when the
@@ -1118,6 +1129,9 @@ def _cmd_create_hero(args: argparse.Namespace) -> int:
         ux.console.print("   check the owner/name spelling (private repos need a reachable URL); "
                          "pin a commit with --sha to skip HEAD resolution")
         return 1
+    dashboard_run_url = f"{dashboard_url}/run/{key}" if dashboard_url else None
+    if dashboard_run_url and getattr(args, "open_dashboard", True):
+        _open_dashboard(dashboard_run_url)
     if verdict == "exists":
         ux.console.print(f"[dim]▶ {key}: run exists — resuming from where it parked[/dim]")
         # A resumed run that already DROPPED can't be driven — re-surface WHY (the recorded gate
@@ -1134,7 +1148,6 @@ def _cmd_create_hero(args: argparse.Namespace) -> int:
         if out.advice:
             ux.console.print(f"   {out.advice}")
         return 1
-    dashboard_run_url = f"{dashboard_url}/run/{key}" if dashboard_url else None
     outcome = ux.drive_run_foreground(run_dir, ctx=_drive_ctx(args), interval=args.interval,
                                       notes_path=runs_dir.parent / "WORKFLOW_NOTES.md")
     ux.summary_panel({key: outcome})
@@ -1614,6 +1627,10 @@ def _add_drive_flags(dp):
                         help="port for the auto-started dashboard (default: 8765)")
         dp.add_argument("--dashboard-host", default="127.0.0.1",
                         help="host for the auto-started dashboard (default: 127.0.0.1)")
+        dp.add_argument("--open-dashboard", dest="open_dashboard", action="store_true", default=True,
+                        help="open the task dashboard in your browser once the run is created (default: on)")
+        dp.add_argument("--no-open-dashboard", dest="open_dashboard", action="store_false",
+                        help="start the dashboard without opening a browser tab")
 
     # ---- hero commands (the README surface) ----
     pcre = sub.add_parser(

src/programsmith/config.py

@@ -53,6 +53,16 @@ class LhConfig(BaseModel):
     openai_api_key: str | None = None
     gemini_api_key: str | None = None      # Google AI Studio key (GEMINI_API_KEY / GOOGLE_API_KEY)
     zai_api_key: str | None = None         # Z.ai key (GLM models)
+
+    # ---- hosted Oddish handoff -------------------------------------------------
+    # ProgramSmith can upload an exported task, launch one hosted trial, and publish the resulting
+    # experiment. The key belongs to the local operator's Oddish account and is stored with the same
+    # owner-only permissions as model credentials. It is never exposed unmasked by the dashboard.
+    oddish_api_key: str | None = None
+    oddish_api_url: str = "https://abundant-ai--api.modal.run"
+    oddish_dashboard_url: str = "https://www.oddish.app"
+    oddish_agent: str = "claude-code"
+    oddish_model: str = "anthropic/claude-sonnet-4-6"
     # Per-TRIAL cost cap handed to the mini-swe solver (`-l`; 0 = disabled). Deliberately 0 by
     # default: the product policy is cost preview + confirm before a sweep, not a silent cap that
     # kills long legitimate trials. Always passed explicitly (mini's own default would cap quietly).
@@ -123,6 +133,11 @@ def load(cls) -> "LhConfig":
         cfg.gemini_api_key = (os.getenv("GEMINI_API_KEY") or os.getenv("GOOGLE_API_KEY")
                               or cfg.gemini_api_key)
         cfg.zai_api_key = os.getenv("ZAI_API_KEY") or cfg.zai_api_key
+        cfg.oddish_api_key = os.getenv("ODDISH_API_KEY") or cfg.oddish_api_key
+        cfg.oddish_api_url = os.getenv("ODDISH_API_URL") or cfg.oddish_api_url
+        cfg.oddish_dashboard_url = (
+            os.getenv("ODDISH_DASHBOARD_URL") or cfg.oddish_dashboard_url
+        )
         if v := os.getenv("PROGRAMSMITH_TRIAL_COST_LIMIT"):
             try:
                 cfg.trial_cost_limit = float(v)
@@ -153,6 +168,7 @@ def mask(v: str | None) -> str | None:
         d["openai_api_key"] = mask(self.openai_api_key)
         d["gemini_api_key"] = mask(self.gemini_api_key)
         d["zai_api_key"] = mask(self.zai_api_key)
+        d["oddish_api_key"] = mask(self.oddish_api_key)
         return d