lcode-cli 0.21.0__tar.gz → 0.23.0__tar.gz

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/CHANGELOG.md

@@ -6,6 +6,34 @@ All notable changes to lcode are documented here. The format follows
 
 ## [Unreleased]
 
+## [0.23.0] - 2026-10-10
+
+### Added
+
+- lcode notices when the model goes round in circles, making the same tool call with the same result
+  again and again, as small models sometimes do. The third time in a request (the second, for a call
+  that failed) the result tells the model so; two more, and lcode ends the request (status `loop`,
+  exit code `4` with `lcode -p`). Results that change, and checks run after an edit, don't count.
+  `repeat_limit` changes the threshold; `0` turns it off. See
+  [When the model repeats itself](https://nasser1941.github.io/lcode/how-it-works/#when-the-model-repeats-itself).
+
+### Fixed
+
+- A tool call Ollama can't parse because its XML is broken (`XML syntax error … unexpected EOF`, from
+  models such as Qwen3.6 that write their calls in XML) is now sent back to the model to make again,
+  like a call with broken JSON, instead of ending the request with an error.
+
+## [0.22.0] - 2026-10-06
+
+### Added
+
+- `lcode mcp add xai`: images, image edits, videos and speech with xAI's Grok Imagine and text to
+  speech APIs, in the cloud and billed to your xAI account. It asks for your xAI API key, and runs
+  `lcode-mcp-xai`, a new MCP server that ships with lcode, with the tools `generate_image`,
+  `edit_image`, `generate_video` (waits for the video and downloads it), `video_status` and
+  `text_to_speech`. Files are saved in `generated/` in the project or where you ask. See
+  [Images, videos and speech with xAI](https://nasser1941.github.io/lcode/mcp/#images-videos-and-speech-with-xai).
+
 ## [0.21.0] - 2026-10-06
 
 ### Added
@@ -480,7 +508,9 @@ First public release.
 - `AGENTS.md` project instructions and `/init` to generate them.
 - One-line installer for Linux and macOS.
 
-[Unreleased]: https://github.com/nasser1941/lcode/compare/v0.21.0...HEAD
+[Unreleased]: https://github.com/nasser1941/lcode/compare/v0.23.0...HEAD
+[0.23.0]: https://github.com/nasser1941/lcode/compare/v0.22.0...v0.23.0
+[0.22.0]: https://github.com/nasser1941/lcode/compare/v0.21.0...v0.22.0
 [0.21.0]: https://github.com/nasser1941/lcode/compare/v0.20.0...v0.21.0
 [0.20.0]: https://github.com/nasser1941/lcode/compare/v0.19.0...v0.20.0
 [0.19.0]: https://github.com/nasser1941/lcode/compare/v0.18.0...v0.19.0

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: lcode-cli
-Version: 0.21.0
+Version: 0.23.0
 Summary: A local-first terminal coding agent powered by open-weight models on your own GPU or Mac, via Ollama.
 Project-URL: Homepage, https://nasser1941.github.io/lcode/
 Project-URL: Documentation, https://nasser1941.github.io/lcode/
@@ -116,7 +116,8 @@ through [Ollama](https://ollama.com), and when it needs current information it c
   (`lcode mcp add atlassian`), image generation with ComfyUI, or any other MCP server. Browser sign-in (OAuth) is built in.
 - **Sees images.** Attach a screenshot or mockup with `@path` and lcode looks at it, using the
   vision part of your model. With ComfyUI (`lcode mcp add comfyui`) it can also generate and edit
-  images with local models such as FLUX, SDXL and Qwen-Image.
+  images with local models such as FLUX, SDXL and Qwen-Image, or, in the cloud, images, videos
+  and speech with xAI's Grok Imagine (`lcode mcp add xai`).
 - **Web search when needed.** Looks up the latest versions, docs and error messages with Ollama web
   search, Brave, Tavily or your own SearXNG, and reads pages as clean text.
 - **Private.** The model runs on your machine, lcode never uploads your files and has no telemetry.

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/README.md

@@ -80,7 +80,8 @@ through [Ollama](https://ollama.com), and when it needs current information it c
   (`lcode mcp add atlassian`), image generation with ComfyUI, or any other MCP server. Browser sign-in (OAuth) is built in.
 - **Sees images.** Attach a screenshot or mockup with `@path` and lcode looks at it, using the
   vision part of your model. With ComfyUI (`lcode mcp add comfyui`) it can also generate and edit
-  images with local models such as FLUX, SDXL and Qwen-Image.
+  images with local models such as FLUX, SDXL and Qwen-Image, or, in the cloud, images, videos
+  and speech with xAI's Grok Imagine (`lcode mcp add xai`).
 - **Web search when needed.** Looks up the latest versions, docs and error messages with Ollama web
   search, Brave, Tavily or your own SearXNG, and reads pages as clean text.
 - **Private.** The model runs on your machine, lcode never uploads your files and has no telemetry.

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/docs/automation.md

@@ -23,7 +23,8 @@ lcode -p "Explain the retry logic" --output stream-json --max-steps 20
 | `--allowed-tools LIST` | Only these tools, e.g. `"read_file,grep,glob,bash"` |
 
 lcode's usual output goes to stderr, so stdout holds only the JSON. The exit code says how it
-went: `0` success, `1` an error, `3` stopped at `--max-steps`, `130` interrupted.
+went: `0` success, `1` an error, `3` stopped at `--max-steps`, `4` stopped because the model
+[kept repeating itself](how-it-works.md#when-the-model-repeats-itself), `130` interrupted.
 
 ### The result
 
@@ -49,7 +50,7 @@ went: `0` success, `1` an error, `3` stopped at `--max-steps`, `130` interrupted
 
 | Field | |
 |---|---|
-| `status` | `success`, `error`, `max_steps` or `interrupted` |
+| `status` | `success`, `error`, `max_steps`, `loop` (the model kept repeating the same call) or `interrupted` |
 | `text` | The model's final answer |
 | `tool_calls` | Every tool call, with its arguments and output (the first 4,000 characters). `error` is true when the call failed or wasn't allowed |
 | `files_changed` | Files the request added (`A`), modified (`M`) or deleted (`D`), shell commands included |

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/docs/configuration.md

@@ -42,6 +42,7 @@ lcode config path                      # print the file location
 | `max_parallel_agents` | `1` | Subagents that run at the same time; more needs `OLLAMA_NUM_PARALLEL` on the Ollama server ([details](agents.md#several-at-once)) |
 | `notify` | `true` | A desktop [notification](usage.md#notifications) when a long request is done or waits for your answer |
 | `notify_after` | `30` | Seconds a request runs before it notifies |
+| `repeat_limit` | `3` | Identical tool calls with identical results before lcode [steps in](how-it-works.md#when-the-model-repeats-itself); `0` turns it off |
 | `checkpoints` | `true` | Save a checkpoint before the model changes files, so [`/undo`](usage.md#undo-and-checkpoints) can restore them |
 
 Besides these settings, `config.toml` can hold [hooks](hooks.md) (`[[hooks]]`) and

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/docs/how-it-works.md

@@ -29,6 +29,25 @@ All tools return text, and failures come back to the model as `Error: …` messa
 crashing the session, so the model can correct itself (for example re-read a file when an edit
 didn't match).
 
+### When the model repeats itself
+
+Small models sometimes fall into a loop: they make a call, get a result, and make the very same call
+again, and again. lcode counts, within a request, the calls that repeat an earlier one exactly: the
+same tool, the same arguments and the same result.
+
+- **At the third one** (the second, when the call failed) the result gets a note: the call returned
+  the same thing several times, so the model should use what it got, try something different, or
+  tell you what's blocking it.
+- **Two more,** and lcode ends the request with
+  `Stopped: the model kept making the same call with the same result`. The JSON output's status is
+  `loop`.
+
+A result that changes isn't a repeat, so a test that's run again after a fix, or a background job
+that's checked while it runs, never adds up. A call that may have changed something, like an edit or
+a shell command that isn't read-only, starts the count again for the other calls: checking the build
+after each edit is fine. Change the threshold with `lcode config set repeat_limit 5`, or turn it off
+with `0`.
+
 ## Subagents
 
 The `agent` tool starts a [subagent](agents.md): another agent loop with its own conversation, the

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/docs/mcp.md

@@ -29,6 +29,7 @@ and find the code for the top one."*
 | `github` | Repositories, issues, pull requests, Actions | a GitHub token (or the GitHub CLI) |
 | `playwright` | A real (headless) browser: open pages, click, fill forms | `npx` |
 | `comfyui` | Generate and edit images with local models (FLUX, SDXL, SD 1.5, Qwen-Image) | `uvx` and [ComfyUI](#images-with-comfyui) running locally |
+| `xai` | Generate and edit images, and generate videos and speech, with xAI's Grok Imagine (paid, in the cloud) | `uvx`, an xAI API key ([below](#images-videos-and-speech-with-xai)) |
 | `context7` | Up-to-date docs and examples for thousands of libraries | nothing (an API key is optional) |
 | `sentry` | Errors, issues, traces and releases | a Sentry account (browser sign-in) |
 | `postgres` | Schemas, read-only queries, query performance | `uvx`, a connection URL |
@@ -137,6 +138,36 @@ size, prompt), which can take several minutes of trial and error. Once an image
 lcode to save that workflow in your project (for example `assets/icon.workflow.json`) and reuse it:
 later images are a single `run_workflow` call.
 
+### Images, videos and speech with xAI
+
+xAI's [Grok Imagine](https://docs.x.ai/developers/model-capabilities/imagine) and
+[text to speech](https://docs.x.ai/developers/models/text-to-speech) APIs make images, videos and
+speech in the cloud: no GPU needed, but each call is billed to your xAI account. lcode ships an MCP
+server for them, `lcode-mcp-xai`.
+
+```bash
+lcode mcp add xai             # asks for your xAI API key (create one at https://console.x.ai)
+```
+
+The key is typed hidden and kept in `~/.config/lcode/mcp.json`, readable only by you. If
+`XAI_API_KEY` is already set in your environment, lcode uses that and doesn't store the key.
+
+| Tool | | Costs |
+|---|---|---|
+| `generate_image` | Images from a description: 1–10 at a time, aspect ratio, 1k or 2k | $0.02–0.05 an image |
+| `edit_image` | Change a local image as described | the same |
+| `generate_video` | A 1–15 second MP4 from a description, optionally starting from a local image, up to 1080p. Waits for it (a minute or more) | $0.02–0.08 a second |
+| `video_status` | Fetch a video that wasn't ready within 10 minutes | |
+| `text_to_speech` | Speech in one of 26 voices and 20 languages, as MP3 or WAV | per character |
+
+Files are saved in `generated/` in the project, or where you ask (*"generate an app icon and save
+it as assets/icon.png"*). xAI has no music generation API.
+
+Your prompts, and the images you edit or animate, go to xAI. lcode asks before each call, since each
+one costs money. To stop asking for some tools, list them under `"allow"` for the server in
+`mcp.json`, for example `"allow": ["generate_image", "text_to_speech"]`. xAI applies its own usage
+policies to what it generates; when it refuses a prompt, lcode passes its message to the model.
+
 ### Metabase
 
 Metabase has its own MCP server inside every instance (Metabase 60 and later), at

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/docs/quickstart.md

@@ -12,7 +12,7 @@ lcode
 
 ```text
 ╭──────────────────────────────────────────────╮
-│ lcode v0.21.0 — local coding agent           │
+│ lcode v0.23.0 — local coding agent           │
 │                                              │
 │ model    lcode-qwen3.6-35b                   │
 │ context  256K tokens                         │

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/pyproject.toml

@@ -49,6 +49,7 @@ Changelog = "https://github.com/nasser1941/lcode/blob/main/CHANGELOG.md"
 lcode = "lcode.cli:main"
 lcode-mcp-valohai = "lcode.servers.valohai:main"
 lcode-mcp-encord = "lcode.servers.encord:main"
+lcode-mcp-xai = "lcode.servers.xai:main"
 
 [dependency-groups]
 dev = ["pytest>=8", "ruff>=0.6"]
@@ -88,6 +89,7 @@ ignore = [
 "src/lcode/action.py" = ["E501"]
 "src/lcode/acp.py" = ["E501"]
 "src/lcode/update.py" = ["E501"]
+"src/lcode/servers/xai.py" = ["E501"]
 
 [tool.pytest.ini_options]
 testpaths = ["tests"]

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/src/lcode/__init__.py

@@ -3,7 +3,7 @@
 From a program: `from lcode import Session` (see lcode.api).
 """
 
-__version__ = "0.21.0"
+__version__ = "0.23.0"
 
 API = ("Session", "Result", "Options", "SetupError")
 

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/src/lcode/acp.py

@@ -48,7 +48,12 @@ KINDS = {
     "bash": "execute", "web_search": "fetch", "web_fetch": "fetch", "agent": "think", "todo_write": "think",
     "present_plan": "switch_mode", "bash_output": "read", "bash_stop": "execute",
 }  # fmt: skip
-STOP = {"success": "end_turn", "max_steps": "max_turn_requests", "interrupted": "cancelled"}
+STOP = {
+    "success": "end_turn",
+    "max_steps": "max_turn_requests",
+    "loop": "max_turn_requests",
+    "interrupted": "cancelled",
+}
 MAX_INLINE = 200_000  # bytes of a file the editor attaches that go into the prompt
 METHOD_NOT_FOUND, INVALID_PARAMS, INTERNAL_ERROR, NOT_FOUND = -32601, -32602, -32603, -32002
 

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/src/lcode/action.py

@@ -297,7 +297,7 @@ def answer(hub: GitHub, task: Task, inputs: Inputs, cwd: Path) -> None:
     )
     with open_session(inputs, cwd) as session:
         result = session.run(prompt)
-        if result.status not in ("success", "max_steps"):
+        if result.status not in ("success", "max_steps", "loop"):
             raise ActionError(result.error or result.status)
         reply = result.text.strip() or "(no answer)"
         changed = git(cwd, "status", "--porcelain").strip()
@@ -336,6 +336,8 @@ def answer(hub: GitHub, task: Task, inputs: Inputs, cwd: Path) -> None:
             reply += f"\n\n{link}"
         if result.status == "max_steps":
             reply += f"\n\n(Stopped after {inputs.max_steps} steps.)"
+        elif result.status == "loop":
+            reply += "\n\n(Stopped: the model kept repeating the same call.)"
         hub.comment(task.number, reply[:MAX_COMMENT] + footer(result))
 
 
@@ -349,7 +351,7 @@ def review(hub: GitHub, task: Task, inputs: Inputs, cwd: Path) -> None:
         prompt = gitflow.review_prompt(session.agent, base)
         session.agent.no_changes = "This is a review, so nothing can be changed: report what should change instead."
         result = session.run(prompt)
-    if result.status not in ("success", "max_steps"):
+    if result.status not in ("success", "max_steps", "loop"):
         raise ActionError(result.error or result.status)
     hub.comment(task.number, f"**lcode review**\n\n{result.text.strip()[:MAX_COMMENT]}{footer(result)}")
 

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/src/lcode/agent.py

@@ -32,6 +32,7 @@ from lcode.mcp import McpManager
 from lcode.ollama import Ollama, OllamaError
 from lcode.permissions import Permissions
 from lcode.render import MarkdownStreamer
+from lcode.repeats import Repeats
 from lcode.sandbox import Sandbox, SandboxError, project_root
 from lcode.tools import (
     LSP_SCHEMA,
@@ -57,7 +58,9 @@ MAX_STEPS_PER_TURN = 150
 EVENT_OUTPUT = 4000  # characters of a tool's output in an event
 # How a tool result starts when the call failed or wasn't allowed (an "error" in events and results).
 REFUSED = ("Error:", "A deny rule", "The user denied", "This needs the user's permission")
-MAX_MALFORMED_CALL_RETRIES = 2  # Ollama rejects tool calls whose arguments aren't valid JSON  # safety cap on tool-call iterations for one request
+MAX_MALFORMED_CALL_RETRIES = 2  # Ollama rejects tool calls whose arguments aren't valid JSON
+# How Ollama reports a tool call it couldn't parse (Qwen3.6's calls are XML)
+MALFORMED_CALL = ("error parsing tool call", "XML syntax error")
 AUTO_COMPACT_RATIO = 0.85  # summarize the history when the context is this full
 PRUNED_ENOUGH = 0.6  # after pruning old tool output, summarize only if the context is still this full
 PROJECT_FILES = ("AGENTS.md", "LCODE.md", "CLAUDE.md")
@@ -216,6 +219,7 @@ class Settings:
     notify: bool = False  # desktop notifications after long requests (lcode.notify)
     notify_after: int = 30  # seconds: a request this long notifies when it's done or waits for an answer
     max_parallel_agents: int = 1
+    repeat_limit: int = 3  # identical calls with identical results before the model is told (lcode.repeats)
 
 
 class Agent:
@@ -267,7 +271,8 @@ class Agent:
         # Decides on a presented plan instead of asking in the terminal: "ask", "auto-edit", or why not.
         self.plan_review: Callable[[str, str], str] | None = None
         self.max_steps = MAX_STEPS_PER_TURN
-        self.turn_status = ""  # how the last request ended: success, max_steps
+        self.turn_status = ""  # how the last request ended: success, max_steps, loop
+        self.repeats = Repeats(settings.repeat_limit)  # notices when the model repeats itself
         self.response = None  # the streaming response, so another thread can abort it
         self.ctx_used = 0
         self.last_speed = 0.0
@@ -809,7 +814,9 @@ class Agent:
         self.turn_status = "running"
         note = planning.NOTE if self.planning() and self.allowed_tools is None else ""
         self.messages.append({"role": "user", "content": self.expand_mentions(user_text) + note})
+        self.repeats.reset()
         malformed = 0
+        looping = ""  # the call the model kept repeating, when lcode stopped the request for it
         for step in range(self.max_steps):
             if self.cancel is not None and self.cancel.is_set():
                 raise KeyboardInterrupt
@@ -834,17 +841,22 @@ class Agent:
                     }
                 )
             except OllamaError as e:
-                if "error parsing tool call" not in str(e) or malformed >= MAX_MALFORMED_CALL_RETRIES:
+                if not any(m in str(e) for m in MALFORMED_CALL) or malformed >= MAX_MALFORMED_CALL_RETRIES:
                     raise
                 malformed += 1
-                reason = str(e).rsplit("err=", 1)[-1].strip() if "err=" in str(e) else "invalid JSON"
+                if "err=" in str(e):
+                    reason = str(e).rsplit("err=", 1)[-1].strip()
+                elif "XML syntax error" in str(e):
+                    reason = str(e).split("XML syntax error", 1)[1].strip(" :") or "invalid XML"
+                else:
+                    reason = "invalid JSON"
                 self.console.print("[yellow]The model wrote a malformed tool call; asking it to try again.[/]")
                 self.messages.append(
                     {
                         "role": "user",
-                        "content": f"[lcode] Your last tool call could not be parsed ({reason}): its arguments "
-                        "must be one complete, valid JSON object. Make the call again. If it writes a file, "
-                        "make sure the whole content is included and properly escaped.",
+                        "content": f"[lcode] Your last tool call could not be parsed ({reason}): it must be "
+                        "complete and well-formed, with every argument closed. Make the call again. If it writes "
+                        "a file, make sure the whole content is included and properly escaped.",
                     }
                 )
                 continue
@@ -892,6 +904,15 @@ class Agent:
                     self.console.print(Text(f"  {result[:300]}", style="red"))
                 elif name in ("read_file", "grep", "glob", "list_dir"):
                     self.console.print(Text(f"  ⎿ {result.count(chr(10)) + 1} line(s)", style="dim"))
+                verdict = self.repeats.check(name, args, result)
+                if verdict.note:
+                    result += verdict.note
+                    what = "stopping the request" if verdict.stop else "told the model"
+                    self.console.print(
+                        Text(f"  ⎿ same call, same result, {verdict.count} times: {what}", style="yellow")
+                    )
+                    if verdict.stop:
+                        looping = self.describe_call(name, args)
                 tool_msg = {"role": "tool", "tool_name": name, "content": result}
                 if call.get("id"):
                     tool_msg["tool_call_id"] = call["id"]
@@ -905,6 +926,27 @@ class Agent:
                         "output": result[:EVENT_OUTPUT] + ("…" if len(result) > EVENT_OUTPUT else ""),
                     }
                 )
+                if looping:
+                    for rest in calls[i + 1 :]:
+                        skipped = {
+                            "role": "tool",
+                            "tool_name": rest.get("function", {}).get("name", ""),
+                            "content": "Not run: lcode stopped the request before this call.",
+                        }
+                        if rest.get("id"):
+                            skipped["tool_call_id"] = rest["id"]
+                        self.messages.append(skipped)
+                    break
+            if looping:
+                self.turn_status = "loop"
+                self.console.print(
+                    Text(
+                        f"Stopped: the model kept making the same call with the same result: {looping}. Tell it "
+                        "what to do differently, or try again.",
+                        style="yellow",
+                    )
+                )
+                return
         else:
             self.turn_status = "max_steps"
             self.console.print(f"[yellow]Stopped after {self.max_steps} steps.[/]")

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/src/lcode/api.py

@@ -35,8 +35,8 @@ if TYPE_CHECKING:
     from lcode.hardware import Hardware
 
 SCHEMA_VERSION = 1
-STATUSES = ("success", "error", "max_steps", "interrupted")
-EXIT_CODES = {"success": 0, "error": 1, "max_steps": 3, "interrupted": 130}
+STATUSES = ("success", "error", "max_steps", "loop", "interrupted")
+EXIT_CODES = {"success": 0, "error": 1, "max_steps": 3, "loop": 4, "interrupted": 130}
 OLLAMA_INSTALL = {
     "linux": "curl -fsSL https://ollama.com/install.sh | sh",
     "darwin": "brew install ollama   (or download it from https://ollama.com/download)",
@@ -69,7 +69,7 @@ class Options:
 class Result:
     """How a request went. `to_json()` is what `lcode -p --output json` prints."""
 
-    status: str  # success, error, max_steps or interrupted
+    status: str  # success, error, max_steps, loop (stopped: the model kept repeating itself) or interrupted
     text: str  # the model's final answer
     session_id: str
     model: str
@@ -194,6 +194,7 @@ def open_agent(cwd: Path, options: Options, console: Console, hw: Hardware | Non
         embed_model=cfg["embed_model"],
         notify=cfg["notify"],
         notify_after=cfg["notify_after"],
+        repeat_limit=cfg["repeat_limit"],
     )
     agent = Agent(ollama, settings, cwd, console=console)
     agent.interactive = options.interactive
@@ -272,7 +273,7 @@ def run_request(agent: Agent, prompt: str) -> Result:
 
     try:
         agent.run_turn(prompt)
-        status = "max_steps" if agent.turn_status == "max_steps" else "success"
+        status = agent.turn_status if agent.turn_status in ("max_steps", "loop") else "success"
     except KeyboardInterrupt:
         status, error = "interrupted", "interrupted before the request was done"
     except ModelError as e:

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/src/lcode/config.py

@@ -59,6 +59,7 @@ SETTINGS: dict[str, tuple[object, type, str]] = {
     "subagents": (True, bool, "let the model hand tasks to subagents that have their own context"),
     "notify": (True, bool, "desktop notification when a long request is done or waits for you"),
     "notify_after": (30, int, "seconds a request runs before it notifies (with notify on)"),
+    "repeat_limit": (3, int, "identical tool calls with identical results before lcode steps in (0: never)"),
     "max_parallel_agents": (1, int, "subagents that may run at the same time (more needs OLLAMA_NUM_PARALLEL)"),
 }
 ENV_OVERRIDES = {

{lcode_cli-0.21.0 → lcode_cli-0.23.0}/src/lcode/mcp/catalog.toml

@@ -281,3 +281,24 @@ secret = true
 var = "VALOHAI_HOST"
 prompt = "Valohai address (Enter for app.valohai.com; for a self-hosted installation, its URL)"
 optional = true
+
+[xai]
+name = "xAI Grok Imagine"
+description = "Generate and edit images, and generate videos and speech, with xAI's Grok Imagine and voice API (paid, in the cloud)"
+homepage = "https://nasser1941.github.io/lcode/mcp/#xai"
+requires = ["uvx"]
+setup = """Uses xAI's API: your prompts, and images you edit or animate, go to xAI, and every call is billed to your xAI
+account (an image $0.02-0.05, video $0.02-0.08 a second, speech per character). Create an API key at https://console.x.ai.
+Files are saved in generated/ in the project. lcode asks before each call; to stop asking for some tools, list them
+under "allow" for this server in mcp.json. xAI has no music generation."""
+
+[xai.server]
+command = "uvx"
+args = ["--isolated", "--python", "3.12", "--from", "lcode-cli", "lcode-mcp-xai"]
+env = { XAI_API_KEY = "${XAI_API_KEY}" }
+timeout = 900  # a video takes minutes
+
+[[xai.inputs]]
+var = "XAI_API_KEY"
+prompt = "xAI API key (create one at https://console.x.ai)"
+secret = true

lcode_cli-0.23.0/src/lcode/repeats.py

@@ -0,0 +1,104 @@
+"""Notice when the model goes round in circles.
+
+Small models sometimes fall into a loop: they make a tool call, get a result, and make the very same call
+again, and again, until the user steps in. Within one request lcode counts the calls that repeat an earlier
+one exactly: the same tool, the same arguments and the same result. At the `limit`-th one (a call that
+failed: one sooner) the result gets a note that tells the model so, and if it still goes on, lcode ends the
+request.
+
+Results that change aren't repeats: a test run again after a fix, a background job read while it runs. And
+a call that may have changed something (a file edit, a shell command that isn't read-only) starts the count
+again for every other call, so checking a build after each edit never adds up. Its own repeats still count:
+running the same failing command over and over is a loop.
+"""
+
+from __future__ import annotations
+
+import hashlib
+import json
+from dataclasses import dataclass
+
+from lcode.permissions import is_read_only
+
+STOP_AFTER_NOTE = 2  # repeats after the note before the request ends
+FAILED = ("Error:", "A deny rule", "The user denied", "This needs the user's permission")
+UNCHANGING = frozenset(
+    {
+        "read_file",
+        "list_dir",
+        "glob",
+        "grep",
+        "view_image",
+        "repo_map",
+        "search_code",
+        "lsp",
+        "web_search",
+        "web_fetch",
+        "bash_output",
+        "todo_write",
+        "skill",
+        "mcp_find_tools",
+    }
+)  # tools that don't change files; anything else might
+
+
+@dataclass
+class Verdict:
+    count: int  # how often this exact call got this exact result in the request
+    note: str = ""  # to add to the tool result
+    stop: bool = False  # end the request
+
+
+class Repeats:
+    def __init__(self, limit: int = 3) -> None:
+        self.limit = limit  # 0: off
+        self.seen: dict[str, int] = {}
+
+    def reset(self) -> None:
+        """A new request."""
+        self.seen.clear()
+
+    def check(self, name: str, args: dict, result: str) -> Verdict:
+        """Count this call and its result."""
+        if self.limit <= 0:
+            return Verdict(1)
+        key = hashlib.sha256(json.dumps([name, args, result], sort_keys=True, default=str).encode()).hexdigest()
+        count = self.seen.get(key, 0) + 1
+        if changes_things(name, args, result):
+            self.seen.clear()
+        self.seen[key] = count
+        failed = result.startswith(FAILED)
+        note_at = max(2, self.limit - 1) if failed else max(2, self.limit)
+        if count < note_at:
+            return Verdict(count)
+        same = "the same error" if failed else "the same result"
+        if count >= note_at + STOP_AFTER_NOTE:
+            return Verdict(
+                count,
+                f"\n\n[lcode] You made this exact call {count} times in this request and got {same} each "
+                "time, so lcode stopped the request here. Next time, don't repeat a call that didn't help: "
+                "change your approach, or tell the user what is blocking you.",
+                stop=True,
+            )
+        hint = (
+            "If it's an edit, read the file again and copy the text exactly. "
+            if failed and name in ("edit_file", "write_file")
+            else ""
+        )
+        return Verdict(
+            count,
+            f"\n\n[lcode] You made this exact call {count} times in this request and got {same} each time. "
+            f"Making it again won't change that. Use what it returned, or try something different. {hint}"
+            "If you're stuck, stop and tell the user what is blocking you.",
+        )
+
+
+def changes_things(name: str, args: dict, result: str) -> bool:
+    """Whether the call may have changed files (or anything else a later call could see)."""
+    if name in UNCHANGING:
+        return False
+    if name in ("edit_file", "write_file"):
+        return not result.startswith(FAILED)
+    if name == "bash":
+        return not is_read_only(str(args.get("command", "")))
+    return True  # other tools (MCP, subagents…) might