lemming-cli 0.4.2__tar.gz → 0.4.3__tar.gz

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/Dockerfile

@@ -11,7 +11,8 @@ RUN curl -fsSL https://deb.nodesource.com/setup_24.x | bash - \
     && rm -rf /var/lib/apt/lists/*
 
 # Install AI coding CLIs (antigravity ships a native binary, not an npm package)
-RUN npm install -g @openai/codex @anthropic-ai/claude-code \
+RUN npm install -g --allow-scripts=opencode-ai \
+      @openai/codex @anthropic-ai/claude-code opencode-ai \
     && curl -fsSL https://antigravity.google/cli/install.sh | bash -s -- --dir /usr/local/bin
 
 # Copy cloudflared and uv binaries directly from their official images

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/PKG-INFO

@@ -1,13 +1,13 @@
 Metadata-Version: 2.5
 Name: lemming-cli
-Version: 0.4.2
+Version: 0.4.3
 Summary: An autonomous, iterative task runner for AI coding agents.
 Project-URL: Homepage, https://github.com/owahltinez/lemming
 Project-URL: Repository, https://github.com/owahltinez/lemming
 Author: Lemming Contributors
 License: MIT
 License-File: LICENSE
-Keywords: agents,ai,aider,antigravity,automation,claude
+Keywords: agents,ai,antigravity,automation,claude,opencode
 Requires-Python: >=3.12
 Requires-Dist: beautifulsoup4>=4.13.3
 Requires-Dist: click>=8.0.0
@@ -38,8 +38,8 @@ human-readable `tasks.yml` file.
   update is recorded in a human-readable `tasks.yml` file. You can step in,
   adjust the roadmap, or swap agents at any time.
 - **Tool Agnostic**: Lemming doesn't care which agent you use. It works
-  out-of-the-box with `agy`, `aider`, `claude`, `codex`, or even your own custom
-  scripts.
+  out-of-the-box with `agy`, `opencode`, `claude`, `codex`, or even your own
+  custom scripts.
 - **Resilient Execution**: With built-in heartbeat monitoring, automatic
   retries, and progress tracking, Lemming handles process crashes and rate
   limits gracefully.
@@ -93,7 +93,7 @@ Start the autonomous loop.
 
 ```bash
 # Run using the project's configured agent (defaults to the first
-# supported runner found on PATH: agy, aider, claude, or codex)
+# supported runner found on PATH: agy, opencode, claude, or codex)
 lemming run
 
 # Flags passed after -- are sent directly to the underlying runner
@@ -150,15 +150,15 @@ inherit the permission prompts of whatever launched it.
 
 ### Teaching Your Agent to Use It
 
-Lemming ships an Agent Skill so that an agent already running — Claude Code,
-Gemini CLI, Cursor — can reach any other agent CLI without knowing how each
-one spells its flags. Install it once:
+Lemming ships an Agent Skill so that an agent already running — Codex, Claude
+Code, Gemini CLI, Cursor — can reach any other agent CLI without knowing how
+each one spells its flags. Install it once:
 
 ```bash
 # Cross-tool location (~/.agents/skills), and report any tool-specific ones
 lemming skill install
 
-# Cover every agent tool detected on this machine
+# Cover detected tools that require their own skill directory too
 lemming skill install --all
 
 # This repository only
@@ -293,7 +293,7 @@ lemming hooks list
 lemming config list
 
 # Persist configuration to tasks.yml
-lemming config set runner aider
+lemming config set runner opencode
 ```
 
 ### Evaluating Prompt Changes
@@ -475,7 +475,10 @@ Lemming uses **fuzzy matching** to automatically inject the correct "YOLO"
 
 - **Antigravity (`agy`)**: Adds `--dangerously-skip-permissions` and exposes
   the project workspace with `--add-dir`
-- **Aider**: Adds `--yes --quiet`
+- **OpenCode**: Runs non-interactively via `opencode run`, adds `--format json`,
+  and, in YOLO mode, adds `--auto`. For its Google provider, Lemming makes an
+  existing `GOOGLE_API_KEY` or `GEMINI_API_KEY` available under OpenCode's
+  native `GOOGLE_GENERATIVE_AI_API_KEY` name.
 - **Claude**: Adds `--dangerously-skip-permissions`
 - **Codex**: Runs non-interactively via `codex exec`, adds `--json`, and, in
   YOLO mode, adds `--dangerously-bypass-approvals-and-sandbox`

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/README.md

@@ -16,8 +16,8 @@ human-readable `tasks.yml` file.
   update is recorded in a human-readable `tasks.yml` file. You can step in,
   adjust the roadmap, or swap agents at any time.
 - **Tool Agnostic**: Lemming doesn't care which agent you use. It works
-  out-of-the-box with `agy`, `aider`, `claude`, `codex`, or even your own custom
-  scripts.
+  out-of-the-box with `agy`, `opencode`, `claude`, `codex`, or even your own
+  custom scripts.
 - **Resilient Execution**: With built-in heartbeat monitoring, automatic
   retries, and progress tracking, Lemming handles process crashes and rate
   limits gracefully.
@@ -71,7 +71,7 @@ Start the autonomous loop.
 
 ```bash
 # Run using the project's configured agent (defaults to the first
-# supported runner found on PATH: agy, aider, claude, or codex)
+# supported runner found on PATH: agy, opencode, claude, or codex)
 lemming run
 
 # Flags passed after -- are sent directly to the underlying runner
@@ -128,15 +128,15 @@ inherit the permission prompts of whatever launched it.
 
 ### Teaching Your Agent to Use It
 
-Lemming ships an Agent Skill so that an agent already running — Claude Code,
-Gemini CLI, Cursor — can reach any other agent CLI without knowing how each
-one spells its flags. Install it once:
+Lemming ships an Agent Skill so that an agent already running — Codex, Claude
+Code, Gemini CLI, Cursor — can reach any other agent CLI without knowing how
+each one spells its flags. Install it once:
 
 ```bash
 # Cross-tool location (~/.agents/skills), and report any tool-specific ones
 lemming skill install
 
-# Cover every agent tool detected on this machine
+# Cover detected tools that require their own skill directory too
 lemming skill install --all
 
 # This repository only
@@ -271,7 +271,7 @@ lemming hooks list
 lemming config list
 
 # Persist configuration to tasks.yml
-lemming config set runner aider
+lemming config set runner opencode
 ```
 
 ### Evaluating Prompt Changes
@@ -453,7 +453,10 @@ Lemming uses **fuzzy matching** to automatically inject the correct "YOLO"
 
 - **Antigravity (`agy`)**: Adds `--dangerously-skip-permissions` and exposes
   the project workspace with `--add-dir`
-- **Aider**: Adds `--yes --quiet`
+- **OpenCode**: Runs non-interactively via `opencode run`, adds `--format json`,
+  and, in YOLO mode, adds `--auto`. For its Google provider, Lemming makes an
+  existing `GOOGLE_API_KEY` or `GEMINI_API_KEY` available under OpenCode's
+  native `GOOGLE_GENERATIVE_AI_API_KEY` name.
 - **Claude**: Adds `--dangerously-skip-permissions`
 - **Codex**: Runs non-interactively via `codex exec`, adds `--json`, and, in
   YOLO mode, adds `--dangerously-bypass-approvals-and-sandbox`

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/SKILL.md

@@ -1,13 +1,13 @@
 ---
 name: lemming
-description: One interface over the AI coding agent CLIs (agy, claude, codex). Delegate a single task to another agent to spare a quota or play to its strengths; run a readability, testing, or UX review over changed files, a branch, or a path; or drive a multi-task project through a roadmap that survives context limits. Use whenever work should run under a different agent than the current one, when reviewing a diff before opening a pull request, or when a job is too large for one agent run.
+description: One interface over the AI coding agent CLIs (agy, claude, codex, opencode). Delegate a single task to another agent to spare a quota or play to its strengths; run a readability, testing, or UX review over changed files, a branch, or a path; or drive a multi-task project through a roadmap that survives context limits. Use whenever work should run under a different agent than the current one, when reviewing a diff before opening a pull request, or when a job is too large for one agent run.
 license: MIT
 ---
 
 # lemming
 
 One interface over several agent CLIs. You do not need to know how `agy`,
-`claude`, or `codex` spell their flags — only `lemming`.
+`claude`, `codex`, or `opencode` spell their flags — only `lemming`.
 
 Use `exec` for one unit of work, and a roadmap for anything larger than a
 single agent run can hold.

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "lemming-cli"
-version = "0.4.2"
+version = "0.4.3"
 description = "An autonomous, iterative task runner for AI coding agents."
 readme = "README.md"
 authors = [
@@ -12,7 +12,7 @@ authors = [
 ]
 license = { text = "MIT" }
 requires-python = ">=3.12"
-keywords = ["ai", "agents", "automation", "antigravity", "claude", "aider"]
+keywords = ["ai", "agents", "automation", "antigravity", "claude", "opencode"]
 dependencies = [
     "beautifulsoup4>=4.13.3",
     "click>=8.0.0",

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/src/lemming/api/config_test.py

@@ -20,7 +20,7 @@ def test_run_loop(client, test_tasks):
 def test_get_runners(client):
     response = client.get("/api/runners")
     assert response.status_code == 200
-    assert response.json() == ["agy", "aider", "claude", "codex"]
+    assert response.json() == ["agy", "opencode", "claude", "codex"]
 
 
 def test_project_goal_isolation(client, test_tasks):

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/src/lemming/cli/config.py

@@ -50,7 +50,7 @@ def config_set(ctx: click.Context, key: str, value: str):
     discard the pinned model.
 
     Examples:
-      lemming config set runner aider
+      lemming config set runner opencode
       lemming config set model gemini-3.6-flash-high
       lemming config set model default
       lemming config set retries 5

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/src/lemming/evals/container.py

@@ -19,6 +19,7 @@ DEFAULT_FORWARD_ENV = (
     "CLAUDE_CODE_OAUTH_TOKEN",
     "GEMINI_API_KEY",
     "GOOGLE_API_KEY",
+    "GOOGLE_GENERATIVE_AI_API_KEY",
 )
 
 # Paths inside the container; the harness mounts host dirs onto these.

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/src/lemming/models.py

@@ -8,7 +8,7 @@ import time
 import pydantic
 
 # Runner CLIs the orchestrator knows how to drive, in default-preference order.
-KNOWN_RUNNERS = ("agy", "aider", "claude", "codex")
+KNOWN_RUNNERS = ("agy", "opencode", "claude", "codex")
 
 
 @functools.cache

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/src/lemming/runner.py

@@ -209,15 +209,21 @@ def build_runner_command(
 
     runner_base = os.path.basename(parts[0])
     is_codex = runner_base.startswith("codex")
+    is_opencode = runner_base.startswith("opencode")
 
-    # Codex's non-interactive interface is the ``exec`` subcommand. Accept an
-    # explicitly configured ``codex exec``/``codex e`` without duplicating the
-    # subcommand, while still making a bare ``codex`` suitable for automation.
+    # Codex and OpenCode expose their non-interactive interfaces as
+    # subcommands. Accept explicitly configured forms without duplicating the
+    # subcommand, while still making a bare executable suitable for automation.
     if is_codex:
         if extra_parts and extra_parts[0] in {"exec", "e"}:
             cmd.append(extra_parts.pop(0))
         else:
             cmd.append("exec")
+    elif is_opencode:
+        if extra_parts and extra_parts[0] == "run":
+            cmd.append(extra_parts.pop(0))
+        else:
+            cmd.append("run")
 
     if not no_defaults:
         if runner_base.startswith("agy"):
@@ -252,11 +258,19 @@ def build_runner_command(
 
             prompt_arg = "--prompt"
         elif runner_base.startswith("aider"):
+            # Backward compatibility for persisted/custom Aider runners. Aider
+            # is no longer advertised as a first-class runner.
             if yolo:
                 cmd.append("--yes")
             if not verbose:
                 cmd.append("--quiet")
             prompt_arg = "--message"
+        elif is_opencode:
+            if yolo:
+                cmd.append("--auto")
+            # Raw JSON events expose completed text, tool activity, and errors
+            # in Lemming's live task log.
+            cmd.extend(["--format", "json"])
         elif runner_base.startswith("claude"):
             if yolo:
                 cmd.append("--dangerously-skip-permissions")
@@ -342,8 +356,19 @@ def extract_error_message(
 
         kind = event.get("type")
         message = None
-        if kind == "error" and isinstance(event.get("message"), str):
-            message = event["message"]
+        if kind == "error":
+            if isinstance(event.get("message"), str):
+                message = event["message"]
+            else:
+                # OpenCode's JSON stream carries the provider error object
+                # under ``error``, usually with its text in ``data.message``.
+                nested = event.get("error")
+                if isinstance(nested, dict):
+                    data = nested.get("data")
+                    if isinstance(data, dict):
+                        message = data.get("message")
+                    if not isinstance(message, str):
+                        message = nested.get("message")
         elif kind == "turn.failed":
             nested = event.get("error")
             if isinstance(nested, dict):
@@ -402,6 +427,16 @@ def _final_message_text(event: dict) -> str | None:
             response = result.get("response")
             return response if isinstance(response, str) else None
 
+    # OpenCode emits completed assistant text as its own event type. Partial
+    # text is not emitted in JSON mode, so the last such event is the closing
+    # response even when tool and step events follow it.
+    if event.get("type") == "text":
+        part = event.get("part")
+        if not isinstance(part, dict) or part.get("type") != "text":
+            return None
+        text = part.get("text")
+        return text if isinstance(text, str) else None
+
     return None
 
 
@@ -493,6 +528,30 @@ def _kill_process_tree(process: subprocess.Popen) -> None:
             pass
 
 
+def _runner_environment(
+    cmd: list[str], tasks_file: pathlib.Path, task_id: str
+) -> dict[str, str]:
+    """Builds the subprocess environment, including runner compatibility.
+
+    OpenCode's native Google provider uses the longer environment variable
+    name, while existing Lemming eval and user environments commonly expose
+    ``GOOGLE_API_KEY`` or ``GEMINI_API_KEY``. Preserve an explicitly
+    configured native value and otherwise make the existing key available
+    under OpenCode's spelling.
+    """
+    env = os.environ.copy()
+    env["LEMMING_PARENT_TASK_ID"] = task_id
+    env["LEMMING_PARENT_TASKS_FILE"] = str(tasks_file.resolve())
+
+    runner_base = os.path.basename(cmd[0]) if cmd else ""
+    if runner_base.startswith("opencode"):
+        google_api_key = env.get("GOOGLE_API_KEY") or env.get("GEMINI_API_KEY")
+        if google_api_key and not env.get("GOOGLE_GENERATIVE_AI_API_KEY"):
+            env["GOOGLE_GENERATIVE_AI_API_KEY"] = google_api_key
+
+    return env
+
+
 def run_with_heartbeat(
     cmd: list[str],
     tasks_file: pathlib.Path,
@@ -545,9 +604,7 @@ def run_with_heartbeat(
 
     # Start the process in a new session so we can kill its entire process
     # tree if needed.
-    env = os.environ.copy()
-    env["LEMMING_PARENT_TASK_ID"] = task_id
-    env["LEMMING_PARENT_TASKS_FILE"] = str(tasks_file.resolve())
+    env = _runner_environment(cmd, tasks_file, task_id)
 
     process = subprocess.Popen(
         cmd,

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/src/lemming/runner_test.py

@@ -80,11 +80,50 @@ def test_build_runner_command_time_limit_ignored_by_other_runners():
 
 
 def test_build_runner_command_aider():
+    """Explicit legacy Aider configurations remain usable."""
     cmd = runner.build_runner_command("aider", "my prompt", yolo=True)
     assert "--yes" in cmd
     assert "--message" in cmd
 
 
+def test_build_runner_command_opencode():
+    cmd = runner.build_runner_command("opencode", "my prompt", yolo=True)
+    assert cmd == [
+        "opencode",
+        "run",
+        "--auto",
+        "--format",
+        "json",
+        "my prompt",
+    ]
+
+
+def test_build_runner_command_opencode_without_yolo():
+    cmd = runner.build_runner_command("opencode", "my prompt", yolo=False)
+    assert cmd == ["opencode", "run", "--format", "json", "my prompt"]
+
+
+def test_build_runner_command_opencode_without_defaults_still_uses_run():
+    cmd = runner.build_runner_command(
+        "opencode", "my prompt", yolo=True, no_defaults=True
+    )
+    assert cmd == ["opencode", "run", "my prompt"]
+
+
+def test_build_runner_command_opencode_does_not_duplicate_explicit_run():
+    cmd = runner.build_runner_command(
+        "opencode run --pure", "my prompt", yolo=False
+    )
+    assert cmd == [
+        "opencode",
+        "run",
+        "--format",
+        "json",
+        "--pure",
+        "my prompt",
+    ]
+
+
 def test_build_runner_command_codex():
     cmd = runner.build_runner_command("codex", "my prompt", yolo=True)
     assert cmd == [
@@ -214,10 +253,13 @@ def test_conflicting_boolean_flag_does_not_consume_next_flag():
 def test_model_is_appended_for_known_runners():
     """A configured model reaches the runner as --model."""
     cmd = runner.build_runner_command(
-        "agy", "prompt", yolo=True, model="gemini-3.6-flash-high"
+        "opencode",
+        "prompt",
+        yolo=True,
+        model="google/gemini-flash-latest",
     )
 
-    assert cmd[cmd.index("--model") + 1] == "gemini-3.6-flash-high"
+    assert cmd[cmd.index("--model") + 1] == "google/gemini-flash-latest"
 
 
 def test_model_defers_to_explicit_runner_string_flag():
@@ -444,6 +486,42 @@ def test_run_with_heartbeat_log_header(tmp_path):
     assert "=" * 80 not in content_2
 
 
+def test_opencode_environment_aliases_google_api_key(tmp_path, monkeypatch):
+    monkeypatch.setenv("GOOGLE_API_KEY", "existing-key")
+    monkeypatch.delenv("GOOGLE_GENERATIVE_AI_API_KEY", raising=False)
+
+    env = runner._runner_environment(
+        ["/usr/local/bin/opencode", "run"], tmp_path / "tasks.yml", "abc"
+    )
+
+    assert env["GOOGLE_GENERATIVE_AI_API_KEY"] == "existing-key"
+
+
+def test_opencode_environment_preserves_native_google_key(
+    tmp_path, monkeypatch
+):
+    monkeypatch.setenv("GOOGLE_API_KEY", "legacy-key")
+    monkeypatch.setenv("GOOGLE_GENERATIVE_AI_API_KEY", "native-key")
+
+    env = runner._runner_environment(
+        ["opencode", "run"], tmp_path / "tasks.yml", "abc"
+    )
+
+    assert env["GOOGLE_GENERATIVE_AI_API_KEY"] == "native-key"
+
+
+def test_opencode_environment_accepts_gemini_api_key(tmp_path, monkeypatch):
+    monkeypatch.delenv("GOOGLE_API_KEY", raising=False)
+    monkeypatch.delenv("GOOGLE_GENERATIVE_AI_API_KEY", raising=False)
+    monkeypatch.setenv("GEMINI_API_KEY", "gemini-key")
+
+    env = runner._runner_environment(
+        ["opencode", "run"], tmp_path / "tasks.yml", "abc"
+    )
+
+    assert env["GOOGLE_GENERATIVE_AI_API_KEY"] == "gemini-key"
+
+
 def test_run_with_heartbeat_records_runner_and_hook_times(tmp_path):
     tasks_file = tmp_path / "tasks.yml"
     task_id = "timed_task"
@@ -814,6 +892,15 @@ def test_extract_error_message_from_turn_failed():
     assert runner.extract_error_message(output) == "quota exhausted"
 
 
+def test_extract_error_message_from_opencode_error_event():
+    output = (
+        '{"type":"error","sessionID":"ses_123","error":'
+        '{"name":"ProviderError","data":{"message":"quota exhausted"}}}\n'
+    )
+
+    assert runner.extract_error_message(output) == "quota exhausted"
+
+
 def test_extract_error_message_ignores_tool_failures():
     """A failed shell command inside the agent is not the runner failing."""
     output = (
@@ -917,6 +1004,15 @@ _AGY_FINAL = (
     '{"event":"result","result":{"status":"SUCCESS",'
     '"response":"All tests pass.\\n","num_turns":1}}\n'
 )
+_OPENCODE_FINAL = (
+    '{"type":"step_start","sessionID":"ses_123","part":'
+    '{"type":"step-start"}}\n'
+    '{"type":"text","sessionID":"ses_123","part":'
+    '{"type":"text","text":"All tests pass.",'
+    '"time":{"start":1,"end":2}}}\n'
+    '{"type":"step_finish","sessionID":"ses_123","part":'
+    '{"type":"step-finish"}}\n'
+)
 
 
 def test_extract_final_message_from_claude_result_event():
@@ -934,6 +1030,11 @@ def test_extract_final_message_from_agy_result_event():
     assert runner.extract_final_message(_AGY_FINAL) == "All tests pass."
 
 
+def test_extract_final_message_from_opencode_text_event():
+    """OpenCode reports completed assistant text in a text-part event."""
+    assert runner.extract_final_message(_OPENCODE_FINAL) == "All tests pass."
+
+
 def test_extract_final_message_skips_non_json_preamble():
     """Codex prints a plain-text notice before its first event."""
     output = "Reading additional input from stdin...\n" + _CODEX_FINAL

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/src/lemming/web/dashboard.spec.js

@@ -55,7 +55,7 @@ test.describe('Dashboard E2E', () => {
       } else if (url.includes('/api/runners')) {
         await route.fulfill({
           contentType: 'application/json',
-          json: ['agy', 'aider'],
+          json: ['agy', 'opencode'],
         });
       } else if (url.includes('/api/hooks')) {
         await route.fulfill({

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/src/lemming/web/favicon.spec.js

@@ -47,7 +47,7 @@ test.describe('Favicon Status Synchronization', () => {
     await page.route('**/api/runners', async (route) => {
       await route.fulfill({
         contentType: 'application/json',
-        json: ['agy', 'aider'],
+        json: ['agy', 'opencode'],
       });
     });
 

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/src/lemming/web/screenshots.spec.js

@@ -58,7 +58,7 @@ const mockTasks = [
     attempts: 0,
     has_runner_log: false,
     pid: null,
-    runner: 'aider',
+    runner: 'opencode',
     parent: 'd4e5f6',
     completed_at: null,
     run_time: null,
@@ -200,7 +200,7 @@ test.describe('Screenshot Generation', () => {
         await page.route('**/api/runners', async (route) => {
           await route.fulfill({
             contentType: 'application/json',
-            json: ['agy', 'aider', 'claude', 'codex'],
+            json: ['agy', 'opencode', 'claude', 'codex'],
           });
         });
         await page.route('**/api/hooks', async (route) => {

{lemming_cli-0.4.2 → lemming_cli-0.4.3}/uv.lock

@@ -222,7 +222,7 @@ wheels = [
 
 [[package]]
 name = "lemming-cli"
-version = "0.4.2"
+version = "0.4.3"
 source = { editable = "." }
 dependencies = [
     { name = "beautifulsoup4" },