kon-coding-agent 0.3.2__tar.gz → 0.3.3__tar.gz

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/AGENTS.md

@@ -7,7 +7,7 @@
 
 ## Testing
 
-- Use `uv run pytest` for testing in general; after edits/writes
+- Use `uv run python -m pytest` for testing in general; after edits/writes
 - If the user asks for e2e tests then run the kon-tmux e2e test if available
 
 ## Committing code
@@ -17,5 +17,5 @@
 
 ## Pushing
 
-- If the user asks you to push code, run these first before doing so: `uv run ruff format .`, `uv run ruff check .`, `uv run pyright .` and `uv run pytest` in parallel (same tool call)
+- If the user asks you to push code, run these first before doing so: `uv run ruff format .`, `uv run ruff check .`, `uv run python -m pyright .` and `uv run python -m pytest` in parallel (same tool call)
 - Only if these all pass without issues should you push otherwise report the warnings/errors back to user and ask for next steps

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/CHANGELOG.md

@@ -30,7 +30,6 @@ All notable changes to this project will be documented in this file.
 - Added tool permission controls with bash safety analysis - @Meltedd.
 - Added popular built-in themes.
 - Added tool previews in approval prompts.
-- Added run token throughput metrics.
 
 ### Changed
 

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kon-coding-agent
-Version: 0.3.2
+Version: 0.3.3
 Summary: Minimal coding agent
 License-File: LICENSE
 Requires-Python: >=3.12
@@ -167,7 +167,7 @@ Here is the full config shape:
 
 ```toml
 [meta]
-config_version = 3
+config_version = 4
 
 [llm]
 default_provider = "openai-codex"
@@ -176,6 +176,10 @@ default_base_url = ""
 default_thinking_level = "high"
 tool_call_idle_timeout_seconds = 180
 
+[llm.auth]
+openai_compat = "auto"
+anthropic_compat = "auto"
+
 [llm.system_prompt]
 git_context = true
 content = """You are an expert coding assistant called Kon.
@@ -248,8 +252,9 @@ Kon is built around a terminal UI that stays simple but practical.
 | --- | --- |
 | File reference | Type `@` to fuzzy-search files and folders in the current project |
 | Path completion | Press **Tab** to complete paths like `./`, `../`, `~`, quoted paths, and absolute paths |
-| Queued prompts | You can submit follow-up prompts while the agent is still running |
-| Queue limit | Up to **5** queued prompts |
+| Queued prompts | Press **Enter** while the agent is running to queue a follow-up prompt |
+| Steer queue | Press **Alt+Enter** to queue a steer message that is processed before normal queued prompts |
+| Queue limit | Up to **5** normal queued prompts and **5** steer messages |
 | Model switching | Use `/model` to switch interactively |
 | Session browsing | Use `/resume` to browse prior sessions |
 
@@ -437,73 +442,12 @@ kon --provider openai --model some-model --api-key "$OPENAI_API_KEY"
 
 ### Local models
 
-Kon works well with local models served through an OpenAI-compatible endpoint.
-
-Example with `llama-server`:
-
-```bash
-./llama-server -m <models-dir>/GLM-4.7-Flash-GGUF/GLM-4.7-Flash-Q4_K_M.gguf \
-  -n 8192 \
-  -c 64000
-
-kon --model zai-org/glm-4.7-flash \
-  --provider openai \
-  --base-url http://localhost:8080/v1 \
-  --api-key ""
-```
+Kon works well with local models served through an OpenAI-compatible endpoint. For one-off launches, you can force unauthenticated local behavior with `--openai-compat-auth none` or `--anthropic-compat-auth none`. To make that persistent across sessions, set `[llm.auth] openai_compat = "auto"|"none"` and/or `anthropic_compat = "auto"|"none"` in `~/.kon/config.toml`.
 
 More notes, tested models, and examples live in [docs/local-models.md](docs/local-models.md).
 
 ---
 
-### Configuration
-
-Kon stores config at:
-
-```text
-~/.kon/config.toml
-```
-
-It is created automatically on first run, and old schemas are migrated forward automatically when needed.
-
-Users are recommended to customize this config based on their model, workflow, safety preferences, and UI taste.
-
-Here is the full config shape:
-
-```toml
-[meta]
-config_version = 3
-
-[llm]
-default_provider = "openai-codex"
-default_model = "gpt-5.4"
-default_base_url = ""
-default_thinking_level = "high"
-tool_call_idle_timeout_seconds = 180
-
-[llm.system_prompt]
-git_context = true
-content = """You are an expert coding assistant called Kon.
-..."""
-
-[compaction]
-on_overflow = "continue"
-buffer_tokens = 20000
-
-[agent]
-max_turns = 500
-default_context_window = 200000
-
-[tools]
-extra = ["web_search", "web_fetch"]
-
-[ui]
-theme = "gruvbox-dark"
-
-[permissions]
-mode = "prompt"
-```
-
 ### Permissions
 
 Kon supports two permission modes:
@@ -526,9 +470,8 @@ Kon depends on a few fast CLI tools for file discovery and search:
 
 - **[`fd`](https://github.com/sharkdp/fd)** - required for fast file discovery
 - **[`ripgrep`](https://github.com/BurntSushi/ripgrep)** - required for fast content search
-- **[`eza`](https://github.com/eza-community/eza)** - optional
 
-If `fd` or `rg` are missing, Kon can download them automatically. `eza` is optional and mainly useful for cleaner, `.gitignore`-aware directory listings.
+If `fd` or `rg` are missing, Kon can download them automatically.
 
 ---
 

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/README.md

@@ -149,7 +149,7 @@ Here is the full config shape:
 
 ```toml
 [meta]
-config_version = 3
+config_version = 4
 
 [llm]
 default_provider = "openai-codex"
@@ -158,6 +158,10 @@ default_base_url = ""
 default_thinking_level = "high"
 tool_call_idle_timeout_seconds = 180
 
+[llm.auth]
+openai_compat = "auto"
+anthropic_compat = "auto"
+
 [llm.system_prompt]
 git_context = true
 content = """You are an expert coding assistant called Kon.
@@ -230,8 +234,9 @@ Kon is built around a terminal UI that stays simple but practical.
 | --- | --- |
 | File reference | Type `@` to fuzzy-search files and folders in the current project |
 | Path completion | Press **Tab** to complete paths like `./`, `../`, `~`, quoted paths, and absolute paths |
-| Queued prompts | You can submit follow-up prompts while the agent is still running |
-| Queue limit | Up to **5** queued prompts |
+| Queued prompts | Press **Enter** while the agent is running to queue a follow-up prompt |
+| Steer queue | Press **Alt+Enter** to queue a steer message that is processed before normal queued prompts |
+| Queue limit | Up to **5** normal queued prompts and **5** steer messages |
 | Model switching | Use `/model` to switch interactively |
 | Session browsing | Use `/resume` to browse prior sessions |
 
@@ -419,73 +424,12 @@ kon --provider openai --model some-model --api-key "$OPENAI_API_KEY"
 
 ### Local models
 
-Kon works well with local models served through an OpenAI-compatible endpoint.
-
-Example with `llama-server`:
-
-```bash
-./llama-server -m <models-dir>/GLM-4.7-Flash-GGUF/GLM-4.7-Flash-Q4_K_M.gguf \
-  -n 8192 \
-  -c 64000
-
-kon --model zai-org/glm-4.7-flash \
-  --provider openai \
-  --base-url http://localhost:8080/v1 \
-  --api-key ""
-```
+Kon works well with local models served through an OpenAI-compatible endpoint. For one-off launches, you can force unauthenticated local behavior with `--openai-compat-auth none` or `--anthropic-compat-auth none`. To make that persistent across sessions, set `[llm.auth] openai_compat = "auto"|"none"` and/or `anthropic_compat = "auto"|"none"` in `~/.kon/config.toml`.
 
 More notes, tested models, and examples live in [docs/local-models.md](docs/local-models.md).
 
 ---
 
-### Configuration
-
-Kon stores config at:
-
-```text
-~/.kon/config.toml
-```
-
-It is created automatically on first run, and old schemas are migrated forward automatically when needed.
-
-Users are recommended to customize this config based on their model, workflow, safety preferences, and UI taste.
-
-Here is the full config shape:
-
-```toml
-[meta]
-config_version = 3
-
-[llm]
-default_provider = "openai-codex"
-default_model = "gpt-5.4"
-default_base_url = ""
-default_thinking_level = "high"
-tool_call_idle_timeout_seconds = 180
-
-[llm.system_prompt]
-git_context = true
-content = """You are an expert coding assistant called Kon.
-..."""
-
-[compaction]
-on_overflow = "continue"
-buffer_tokens = 20000
-
-[agent]
-max_turns = 500
-default_context_window = 200000
-
-[tools]
-extra = ["web_search", "web_fetch"]
-
-[ui]
-theme = "gruvbox-dark"
-
-[permissions]
-mode = "prompt"
-```
-
 ### Permissions
 
 Kon supports two permission modes:
@@ -508,9 +452,8 @@ Kon depends on a few fast CLI tools for file discovery and search:
 
 - **[`fd`](https://github.com/sharkdp/fd)** - required for fast file discovery
 - **[`ripgrep`](https://github.com/BurntSushi/ripgrep)** - required for fast content search
-- **[`eza`](https://github.com/eza-community/eza)** - optional
 
-If `fd` or `rg` are missing, Kon can download them automatically. `eza` is optional and mainly useful for cleaner, `.gitignore`-aware directory listings.
+If `fd` or `rg` are missing, Kon can download them automatically.
 
 ---
 

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/docs/local-models.md

@@ -20,9 +20,23 @@ Run a local model using llama-server with the following command:
   -c 65536
 ```
 
-Then start kon:
+Then start Kon for a one-off local session:
 
 ```bash
-kon --model unsloth/Qwen3.5-9B-GGUF --provider openai --base-url http://localhost:5000/v1 --api-key ""
+kon --model unsloth/Qwen3.5-9B-GGUF --provider openai \
+  --base-url http://localhost:5000/v1 \
+  --openai-compat-auth none
+```
+
+If this is your default setup, put it in `~/.kon/config.toml` instead:
+
+```toml
+[llm]
+default_provider = "openai"
+default_model = "unsloth/Qwen3.5-9B-GGUF"
+default_base_url = "http://localhost:5000/v1"
+
+[llm.auth]
+openai_compat = "auto" # or "none" to always inject a placeholder key
 ```
 

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/pyproject.toml

@@ -14,7 +14,7 @@ default = true
 
 [project]
 name = "kon-coding-agent"
-version = "0.3.2"
+version = "0.3.3"
 description = "Minimal coding agent"
 readme = "README.md"
 requires-python = ">=3.12"
@@ -36,6 +36,7 @@ dev = [
  "pyright>=1.1.408",
  "pytest>=9.0.2",
  "pytest-asyncio>=1.3.0",
+ "ruff>=0.15.8",
  "twine>=5.1.1",
 ]
 

kon_coding_agent-0.3.3/src/kon/async_utils.py

@@ -0,0 +1,40 @@
+import asyncio
+from contextlib import suppress
+from typing import Any
+
+
+class OperationCancelledError(Exception):
+    pass
+
+
+async def cancel_and_await(task: asyncio.Future[Any]) -> None:
+    if task.done():
+        return
+    task.cancel()
+    with suppress(asyncio.CancelledError):
+        await task
+
+
+async def await_or_cancel[T](work: asyncio.Future[T], cancel_event: asyncio.Event | None) -> T:
+    if not cancel_event:
+        return await work
+
+    if cancel_event.is_set():
+        await cancel_and_await(work)
+        raise OperationCancelledError
+
+    cancel = asyncio.create_task(cancel_event.wait())
+    try:
+        done, pending = await asyncio.wait({work, cancel}, return_when=asyncio.FIRST_COMPLETED)
+
+        for task in pending:
+            await cancel_and_await(task)
+
+        if cancel in done and cancel_event.is_set():
+            if not work.done():
+                await cancel_and_await(work)
+            raise OperationCancelledError
+
+        return work.result()
+    finally:
+        await cancel_and_await(cancel)

kon_coding_agent-0.3.3/src/kon/builtin_skills/init/SKILL.md

@@ -0,0 +1,73 @@
+---
+name: init
+description: Create or update AGENTS.md for this repository
+register_cmd: true
+cmd_info: Guided AGENTS.md setup
+---
+
+Create or update `AGENTS.md` for this repository.
+
+The goal is a compact instruction file that helps future Kon sessions avoid mistakes and ramp up quickly. Every line should answer: "Would an agent likely miss this without help?" If not, leave it out.
+
+User-provided focus or constraints (honor these):
+$ARGUMENTS
+
+## How to investigate
+
+Read the highest-value sources first:
+- `README*`, root manifests, workspace config, lockfiles
+- build, test, lint, formatter, typecheck, and codegen config
+- CI workflows and pre-commit / task runner config
+- existing instruction files (`AGENTS.md`, `CLAUDE.md`, `.cursor/rules/`, `.cursorrules`, `.github/copilot-instructions.md`)
+- repo-local Kon config and conventions
+
+If architecture is still unclear after reading config and docs, inspect a small number of representative code files to find the real entrypoints, package boundaries, and execution flow. Prefer reading the files that explain how the system is wired together over random leaf files.
+
+Prefer executable sources of truth over prose. If docs conflict with config or scripts, trust the executable source and only keep what you can verify.
+
+## What to extract
+
+Look for the highest-signal facts for an agent working in this repo:
+- exact developer commands, especially non-obvious ones
+- how to run a single test, a single package, or a focused verification step
+- required command order when it matters, such as `lint -> typecheck -> test`
+- monorepo or multi-package boundaries, ownership of major directories, and the real app/library entrypoints
+- framework or toolchain quirks: generated code, migrations, codegen, build artifacts, special env loading, dev servers, infra deploy flow
+- repo-specific style or workflow conventions that differ from defaults
+- testing quirks: fixtures, integration test prerequisites, snapshot workflows, required services, flaky or expensive suites
+- important constraints from existing instruction files worth preserving
+
+Good `AGENTS.md` content is usually hard-earned context that took reading multiple files to infer.
+
+## Questions
+
+Only ask the user questions if the repo cannot answer something important. Use the question tool for one short batch at most.
+
+Good questions:
+- undocumented team conventions
+- branch / PR / release expectations
+- missing setup or test prerequisites that are known but not written down
+
+Do not ask about anything the repo already makes clear.
+
+## Writing rules
+
+Include only high-signal, repo-specific guidance such as:
+- exact commands and shortcuts the agent would otherwise guess wrong
+- architecture notes that are not obvious from filenames
+- conventions that differ from language or framework defaults
+- setup requirements, environment quirks, and operational gotchas
+- references to existing instruction sources that matter
+
+Exclude:
+- generic software advice
+- long tutorials or exhaustive file trees
+- obvious language conventions
+- speculative claims or anything you could not verify
+- content better stored in another file referenced via Kon config or another instruction file
+
+When in doubt, omit.
+
+Prefer short sections and bullets. If the repo is simple, keep the file simple. If the repo is large, summarize the few structural facts that actually change how an agent should work.
+
+If `AGENTS.md` already exists, improve it in place rather than rewriting blindly. Preserve verified useful guidance, delete fluff or stale claims, and reconcile it with the current codebase.

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/src/kon/config.py

@@ -18,6 +18,7 @@ from .themes import ColorsConfig, get_theme, get_theme_ids
 CONFIG_DIR_NAME: str = ".kon"
 
 OnOverflowMode = Literal["continue", "pause"]
+AuthMode = Literal["auto", "required", "none"]
 
 
 # =================================================================================================
@@ -63,6 +64,11 @@ class SystemPromptConfig(BaseModel):
     git_context: bool = False
 
 
+class AuthConfig(BaseModel):
+    openai_compat: AuthMode = "auto"
+    anthropic_compat: AuthMode = "auto"
+
+
 class LLMConfig(BaseModel):
     default_provider: str
     default_model: str
@@ -70,6 +76,7 @@ class LLMConfig(BaseModel):
     default_thinking_level: str
     system_prompt: SystemPromptConfig
     tool_call_idle_timeout_seconds: float = 180
+    auth: AuthConfig = AuthConfig()
 
 
 class CompactionConfig(BaseModel):
@@ -120,10 +127,6 @@ class _BinariesConfig:
     def fd(self) -> bool:
         return "fd" in self._binaries
 
-    @property
-    def eza(self) -> bool:
-        return "eza" in self._binaries
-
 
 class Config:
     def __init__(self, data: dict[str, Any]) -> None:
@@ -227,7 +230,7 @@ def consume_config_warnings() -> list[str]:
 
 
 def _detect_available_binaries() -> set[str]:
-    binaries = {"rg", "fd", "eza"}
+    binaries = {"rg", "fd"}
     available = set()
     bin_dir = Path.home() / CONFIG_DIR_NAME / "bin"
 
@@ -286,6 +289,29 @@ def _migrate_v2_to_v3(data: dict[str, Any]) -> dict[str, Any]:
     return migrated
 
 
+def _migrate_v3_to_v4(data: dict[str, Any]) -> dict[str, Any]:
+    migrated = Config._apply_legacy_key_shims(data)
+    llm = migrated.get("llm")
+    if not isinstance(llm, dict):
+        llm = {}
+        migrated["llm"] = llm
+
+    auth = llm.get("auth")
+    if not isinstance(auth, dict):
+        auth = {}
+        llm["auth"] = auth
+
+    auth.setdefault("openai_compat", "auto")
+    auth.setdefault("anthropic_compat", "auto")
+
+    meta = migrated.get("meta")
+    if not isinstance(meta, dict):
+        migrated["meta"] = {"config_version": 4}
+    else:
+        meta["config_version"] = 4
+    return migrated
+
+
 def _migrate_config_data(data: dict[str, Any]) -> tuple[dict[str, Any], int, int, bool]:
     original = deepcopy(data)
     current_version = _get_config_version(original)
@@ -304,6 +330,10 @@ def _migrate_config_data(data: dict[str, Any]) -> tuple[dict[str, Any], int, int
             migrated = _migrate_v2_to_v3(migrated)
             current_version = 3
             continue
+        if current_version == 3:
+            migrated = _migrate_v3_to_v4(migrated)
+            current_version = 4
+            continue
         break
 
     migrated_version = _get_config_version(migrated)

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/src/kon/context/skills.py

@@ -12,6 +12,7 @@ Discovery locations:
 import os
 import re
 from dataclasses import dataclass
+from importlib import resources
 from pathlib import Path
 from typing import Any
 
@@ -47,6 +48,7 @@ class Skill:
     description: str
     register_cmd: bool = False
     cmd_info: str = ""
+    bundled: bool = False
 
 
 @dataclass
@@ -228,6 +230,66 @@ def load_skills(cwd: str | None = None) -> LoadSkillsResult:
     return LoadSkillsResult(skills=list(skill_map.values()), warnings=all_warnings)
 
 
+def load_builtin_cmd_skills() -> LoadSkillsResult:
+    try:
+        builtin_resource = resources.files("kon").joinpath("builtin_skills")
+        with resources.as_file(builtin_resource) as builtin_root:
+            result = _load_skills_from_dir(builtin_root)
+    except Exception:
+        return LoadSkillsResult(skills=[], warnings=[])
+    return LoadSkillsResult(
+        skills=[
+            Skill(
+                path=skill.path,
+                name=skill.name,
+                description=skill.description,
+                register_cmd=skill.register_cmd,
+                cmd_info=skill.cmd_info,
+                bundled=True,
+            )
+            for skill in result.skills
+        ],
+        warnings=result.warnings,
+    )
+
+
+def strip_frontmatter(content: str) -> str:
+    if not content.startswith("---"):
+        return content.strip()
+    end_match = re.search(r"\n---\s*\n", content[3:])
+    if not end_match:
+        return content.strip()
+    return content[end_match.end() + 3 :].strip()
+
+
+def render_skill_prompt(skill: Skill, query: str) -> str:
+    try:
+        content = Path(skill.path).read_text(encoding="utf-8")
+    except Exception:
+        return _build_fallback_skill_prompt(skill.description, query)
+    template = strip_frontmatter(content)
+    rendered = template.replace("$ARGUMENTS", query)
+    return rendered.strip()
+
+
+def _build_fallback_skill_prompt(description: str, query: str) -> str:
+    query = query.strip()
+    if not query:
+        return description
+    return f"{description}\n\n{query}"
+
+
+def merge_registered_skills(primary: list[Skill], secondary: list[Skill]) -> list[Skill]:
+    seen = {skill.name for skill in primary}
+    merged = list(primary)
+    for skill in secondary:
+        if skill.name in seen:
+            continue
+        merged.append(skill)
+        seen.add(skill.name)
+    return merged
+
+
 def formatted_skills(skills: list[Skill]) -> str:
     if not skills:
         return ""

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/src/kon/core/types.py

@@ -12,6 +12,7 @@ class StopReason(StrEnum):
     TOOL_USE = "tool_use"
     ERROR = "error"
     INTERRUPTED = "interrupted"
+    STEER = "steer"
 
 
 class Usage(BaseModel):

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/src/kon/defaults/config.toml

@@ -1,5 +1,5 @@
 [meta]
-config_version = 3
+config_version = 4
 
 [llm]
 default_provider = "openai-codex" # "zhipu", "github-copilot", "openai-codex"
@@ -12,6 +12,14 @@ default_thinking_level = "high"
 # Helps prevent stalled tool executions from hanging the agent loop.
 tool_call_idle_timeout_seconds = 180
 
+[llm.auth]
+# Auth policy for OpenAI-compatible and Anthropic-compatible endpoints.
+# "auto" injects a placeholder key for local/self-hosted endpoints when no key is set.
+# "required" keeps strict API key requirements.
+# "none" always injects a placeholder key for that protocol family.
+openai_compat = "auto" # "auto", "required", or "none"
+anthropic_compat = "auto" # "auto", "required", or "none"
+
 [llm.system_prompt]
 git_context = true
 content = """You are an expert coding assistant called Kon. You help users by reading, searching, executing commands, editing code, and writing new files.

{kon_coding_agent-0.3.2 → kon_coding_agent-0.3.3}/src/kon/events.py

@@ -41,7 +41,6 @@ class TurnEndEvent:
     assistant_message: AssistantMessage | None = None
     tool_results: list[ToolResultMessage] = field(default_factory=list)
     stop_reason: StopReason = StopReason.STOP
-    generation_seconds: float | None = None
     tool_call_count: int = 0