mycode-sdk 0.5.2__tar.gz → 0.5.3__tar.gz

{mycode_sdk-0.5.2 → mycode_sdk-0.5.3}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: mycode-sdk
-Version: 0.5.2
-Summary: Multi-turn tool-calling agent runtime for embedding the mycode agent loop.
+Version: 0.5.3
+Summary: Lightweight Python SDK for building AI agents.
 Project-URL: Homepage, https://github.com/legibet/mycode
 Project-URL: Repository, https://github.com/legibet/mycode
 Project-URL: Issues, https://github.com/legibet/mycode/issues
@@ -25,7 +25,7 @@ Description-Content-Type: text/markdown
 
 # mycode-sdk
 
-Lightweight Python SDK for building agents.
+Lightweight Python SDK for building AI agents.
 
 ## Install
 
@@ -58,18 +58,42 @@ async def main() -> None:
 asyncio.run(main())
 ```
 
-`Agent(...)` infers the provider from the model id and defaults `cwd` to the current working directory. No tools are registered unless you pass `tools=[...]`, and nothing is persisted unless you pass `session_dir=`.
+`Agent(...)` infers the provider from the model id. No tools are registered unless you pass `tools=[...]`.
 
-For a synchronous call, use `run()` — it collects the stream into a `RunResult`:
+For a simple synchronous call, use `run()`:
 
 ```python
 result = agent.run("Read pyproject.toml and tell me the project name.")
 print(result.text)
 ```
 
-Call `achat` or `run` again on the same `Agent` to continue the conversation — history accumulates in `agent.messages`.
+## Multi-turn conversations
 
-To persist across processes, pass `session_dir` as the root directory; each `session_id` becomes a subdirectory. Reconstruct with the same `(session_dir, session_id)` to resume.
+Call `achat()` or `run()` again on the same `Agent` — history accumulates automatically:
+
+```python
+agent = Agent(model="claude-sonnet-4-6", api_key="...")
+
+agent.run("What is 2 + 2?")
+agent.run("Now multiply that by 10.")    # remembers the earlier answer
+```
+
+## Saving sessions
+
+Pass `session_dir` to persist the conversation to disk. Each session lives in a subdirectory named by `session_id`:
+
+```python
+from pathlib import Path
+
+agent = Agent(
+    model="claude-sonnet-4-6",
+    api_key="...",
+    session_dir=Path("./chats"),
+    session_id="my-chat",
+)
+```
+
+Construct another `Agent` with the same `(session_dir, session_id)` later to resume the conversation — the history is loaded automatically.
 
 ## Built-in tools
 
@@ -77,33 +101,42 @@ To persist across processes, pass `session_dir` as the root directory; each `ses
 from mycode import read_tool, write_tool, edit_tool, bash_tool
 ```
 
-Only `bash_tool` streams incremental output as `tool_output` events; the others return a single result.
+Four tools for reading, writing, editing files, and running shell commands. Opt in by passing them to `tools=[...]`.
 
 ## Custom tools
 
-`@tool` wraps a sync or `async def` Python function as a `ToolSpec`. Parameter type hints become the JSON schema sent to the provider.
-
-Annotate the first parameter as `ToolContext` to have the context injected. Use `ctx.read / ctx.write / ctx.edit / ctx.bash` to invoke the built-ins, or `ctx.call(name, args)` for any registered tool by name.
+Decorate any function with `@tool`. Parameter type hints become the JSON schema sent to the provider; the docstring becomes the description:
 
 ```python
-from mycode import Agent, ToolContext, read_tool, tool
+from mycode import Agent, tool
 
 
 @tool
-def summarize_file(ctx: ToolContext, path: str) -> str:
-    """Return the first line of a text file."""
+def greet(name: str) -> str:
+    """Return a friendly greeting."""
 
-    result = ctx.read(path)
-    return result.output.splitlines()[0] if result.output else ""
+    return f"hello, {name}"
 
 
 agent = Agent(
     model="claude-sonnet-4-6",
-    api_key="YOUR_API_KEY",
-    tools=[read_tool, summarize_file],
+    api_key="...",
+    tools=[greet],
 )
 ```
 
-A bare `str` return becomes the tool `output`; any other JSON-serializable value is dumped to JSON. For finer control, return a `ToolExecutionResult` to set `output`, `content` (multimodal blocks such as images), `metadata` (structured UI data), and `is_error` independently.
+To call a built-in tool from inside your own, type the first parameter as `ToolContext`:
+
+```python
+from mycode import ToolContext, tool
+
+
+@tool
+def summarize_file(ctx: ToolContext, path: str) -> str:
+    """Return the first line of a text file."""
+
+    result = ctx.read(path)
+    return result.output.splitlines()[0] if result.output else ""
+```
 
-See [docs/sdk.md](../docs/sdk.md) for the event stream, cancellation, session rules, and the full `Agent` / `@tool` reference.
+See [docs/sdk.md](../docs/sdk.md) for the event stream, cancellation, sessions, and the full `Agent` / `@tool` reference.

mycode_sdk-0.5.3/README.md

@@ -0,0 +1,117 @@
+# mycode-sdk
+
+Lightweight Python SDK for building AI agents.
+
+## Install
+
+```bash
+uv add mycode-sdk
+# or
+pip install mycode-sdk
+```
+
+## Quick start
+
+```python
+import asyncio
+
+from mycode import Agent, bash_tool, read_tool
+
+
+async def main() -> None:
+    agent = Agent(
+        model="claude-sonnet-4-6",
+        api_key="YOUR_API_KEY",
+        tools=[read_tool, bash_tool],
+    )
+
+    async for event in agent.achat("Read pyproject.toml and tell me the project name."):
+        if event.type == "text":
+            print(event.data["delta"], end="", flush=True)
+
+
+asyncio.run(main())
+```
+
+`Agent(...)` infers the provider from the model id. No tools are registered unless you pass `tools=[...]`.
+
+For a simple synchronous call, use `run()`:
+
+```python
+result = agent.run("Read pyproject.toml and tell me the project name.")
+print(result.text)
+```
+
+## Multi-turn conversations
+
+Call `achat()` or `run()` again on the same `Agent` — history accumulates automatically:
+
+```python
+agent = Agent(model="claude-sonnet-4-6", api_key="...")
+
+agent.run("What is 2 + 2?")
+agent.run("Now multiply that by 10.")    # remembers the earlier answer
+```
+
+## Saving sessions
+
+Pass `session_dir` to persist the conversation to disk. Each session lives in a subdirectory named by `session_id`:
+
+```python
+from pathlib import Path
+
+agent = Agent(
+    model="claude-sonnet-4-6",
+    api_key="...",
+    session_dir=Path("./chats"),
+    session_id="my-chat",
+)
+```
+
+Construct another `Agent` with the same `(session_dir, session_id)` later to resume the conversation — the history is loaded automatically.
+
+## Built-in tools
+
+```python
+from mycode import read_tool, write_tool, edit_tool, bash_tool
+```
+
+Four tools for reading, writing, editing files, and running shell commands. Opt in by passing them to `tools=[...]`.
+
+## Custom tools
+
+Decorate any function with `@tool`. Parameter type hints become the JSON schema sent to the provider; the docstring becomes the description:
+
+```python
+from mycode import Agent, tool
+
+
+@tool
+def greet(name: str) -> str:
+    """Return a friendly greeting."""
+
+    return f"hello, {name}"
+
+
+agent = Agent(
+    model="claude-sonnet-4-6",
+    api_key="...",
+    tools=[greet],
+)
+```
+
+To call a built-in tool from inside your own, type the first parameter as `ToolContext`:
+
+```python
+from mycode import ToolContext, tool
+
+
+@tool
+def summarize_file(ctx: ToolContext, path: str) -> str:
+    """Return the first line of a text file."""
+
+    result = ctx.read(path)
+    return result.output.splitlines()[0] if result.output else ""
+```
+
+See [docs/sdk.md](../docs/sdk.md) for the event stream, cancellation, sessions, and the full `Agent` / `@tool` reference.

{mycode_sdk-0.5.2 → mycode_sdk-0.5.3}/pyproject.toml

@@ -4,8 +4,8 @@ build-backend = "hatchling.build"
 
 [project]
 name = "mycode-sdk"
-version = "0.5.2"
-description = "Multi-turn tool-calling agent runtime for embedding the mycode agent loop."
+version = "0.5.3"
+description = "Lightweight Python SDK for building AI agents."
 readme = "README.md"
 requires-python = ">=3.12"
 license = "MIT"

{mycode_sdk-0.5.2 → mycode_sdk-0.5.3}/src/mycode/models.py

@@ -12,16 +12,14 @@ from mycode.utils import as_bool, as_int
 
 _MODELS_CATALOG_PATH = Path(__file__).with_name("models_catalog.json")
 
-# Catalogs consulted only for capability bits (context window, image / pdf
-# support, …) when the requested provider has no entry for the model. They
-# are NOT registered providers; the metadata returned from a fallback hit is
-# always attributed to a real provider type the caller already has in hand.
-_FALLBACK_CAPABILITY_CATALOGS: tuple[str, ...] = ("aihubmix",)
-
 
 @dataclass(frozen=True)
 class ModelMetadata:
-    """Normalized metadata used by provider resolution."""
+    """Model metadata for the requested provider/model.
+
+    ``provider`` and ``model`` keep the original query identity. Other fields
+    may come from a fallback catalog entry.
+    """
 
     provider: str
     model: str
@@ -91,71 +89,74 @@ def lookup_model_metadata(
     provider_type: str | None,
     model: str | None,
 ) -> ModelMetadata | None:
-    """Resolve metadata for one provider type and model.
+    """Return model metadata for a provider/model request.
 
-    Lookup tiers, in order:
+    Catalog lookup order:
 
-    1. Exact ``(provider_type, model)`` entry.
-    2. Canonical provider inferred from the model id prefix (``gpt-`` →
-       ``openai``, ``claude-`` → ``anthropic``, …) when the requested
-       provider had no hit.
-    3. Capability fallback from a secondary catalog (currently
-       ``aihubmix``), attributed to the caller's real provider — never
-       to the secondary catalog id.
+    1. Requested provider and requested model.
+    2. Inferred provider and unprefixed model name.
+    3. Unique OpenRouter model whose suffix matches the model name.
     """
 
-    raw = (model or "").strip()
-    if not raw:
+    requested_model = (model or "").strip()
+    if not provider_type or not requested_model:
         return None
     catalog = load_models_catalog()
     if not catalog:
         return None
 
-    bare = raw.split("/", 1)[1].strip() if "/" in raw else raw
-    inferred = infer_provider_from_model(bare)
+    model_name = requested_model.split("/", 1)[1].strip() if "/" in requested_model else requested_model
+    catalog_entry = _get_catalog_entry(catalog, provider_type, requested_model)
 
-    if provider_type:
-        hit = _match(catalog, lookup=provider_type, model_id=raw, attributed=provider_type)
-        if hit is not None:
-            return hit
+    inferred_provider = infer_provider_from_model(model_name)
+    if catalog_entry is None and inferred_provider and inferred_provider != provider_type:
+        catalog_entry = _get_catalog_entry(catalog, inferred_provider, model_name)
 
-    if inferred and inferred != provider_type:
-        hit = _match(catalog, lookup=inferred, model_id=bare, attributed=inferred)
-        if hit is not None:
-            return hit
+    if catalog_entry is None:
+        catalog_entry = _get_openrouter_suffix_entry(catalog, model_name)
 
-    attributed = provider_type or inferred
-    if attributed is None:
+    if catalog_entry is None:
         return None
-    for source in _FALLBACK_CAPABILITY_CATALOGS:
-        hit = _match(catalog, lookup=source, model_id=bare, attributed=attributed)
-        if hit is not None:
-            return hit
 
-    return None
+    return ModelMetadata(
+        provider=provider_type,
+        model=requested_model,
+        context_window=as_int(catalog_entry.get("context_window")),
+        max_output_tokens=as_int(catalog_entry.get("max_output_tokens")),
+        supports_reasoning=as_bool(catalog_entry.get("supports_reasoning")),
+        supports_image_input=as_bool(catalog_entry.get("supports_image_input")),
+        supports_pdf_input=as_bool(catalog_entry.get("supports_pdf_input")),
+    )
 
 
-def _match(
+def _get_catalog_entry(
     catalog: dict[str, Any],
-    *,
-    lookup: str,
+    provider: str,
     model_id: str,
-    attributed: str,
-) -> ModelMetadata | None:
-    """Look up one model in a catalog section and build metadata if present."""
+) -> dict[str, Any] | None:
+    """Return a catalog entry for provider/model_id."""
 
-    section = catalog.get(lookup)
+    section = catalog.get(provider)
     if not isinstance(section, dict):
         return None
-    raw = section.get(model_id)
-    if not isinstance(raw, dict):
+    catalog_entry = section.get(model_id)
+    return catalog_entry if isinstance(catalog_entry, dict) else None
+
+
+def _get_openrouter_suffix_entry(catalog: dict[str, Any], model_name: str) -> dict[str, Any] | None:
+    """Return an OpenRouter entry with a unique matching model suffix."""
+
+    openrouter = catalog.get("openrouter")
+    if not isinstance(openrouter, dict):
         return None
-    return ModelMetadata(
-        provider=attributed,
-        model=model_id,
-        context_window=as_int(raw.get("context_window")),
-        max_output_tokens=as_int(raw.get("max_output_tokens")),
-        supports_reasoning=as_bool(raw.get("supports_reasoning")),
-        supports_image_input=as_bool(raw.get("supports_image_input")),
-        supports_pdf_input=as_bool(raw.get("supports_pdf_input")),
-    )
+
+    match: dict[str, Any] | None = None
+    for model_id, catalog_entry in openrouter.items():
+        if not isinstance(model_id, str) or "/" not in model_id or not isinstance(catalog_entry, dict):
+            continue
+        if model_id.split("/", 1)[1].strip() != model_name:
+            continue
+        if match is not None:
+            return None
+        match = catalog_entry
+    return match

{mycode_sdk-0.5.2 → mycode_sdk-0.5.3}/src/mycode/models_catalog.json

@@ -1,370 +1,4 @@
 {
-  "aihubmix": {
-    "Kimi-K2-0905": {
-      "context_window": 262144,
-      "max_output_tokens": 262144,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "claude-haiku-4-5": {
-      "context_window": 200000,
-      "max_output_tokens": 64000,
-      "supports_image_input": true,
-      "supports_pdf_input": true,
-      "supports_reasoning": true
-    },
-    "claude-opus-4-1": {
-      "context_window": 200000,
-      "max_output_tokens": 32000,
-      "supports_image_input": true,
-      "supports_pdf_input": true,
-      "supports_reasoning": true
-    },
-    "claude-opus-4-5": {
-      "context_window": 200000,
-      "max_output_tokens": 32000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "claude-opus-4-6": {
-      "context_window": 200000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": true,
-      "supports_reasoning": true
-    },
-    "claude-opus-4-6-think": {
-      "context_window": 200000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": true,
-      "supports_reasoning": true
-    },
-    "claude-sonnet-4-5": {
-      "context_window": 200000,
-      "max_output_tokens": 64000,
-      "supports_image_input": true,
-      "supports_pdf_input": true,
-      "supports_reasoning": true
-    },
-    "claude-sonnet-4-6": {
-      "context_window": 200000,
-      "max_output_tokens": 64000,
-      "supports_image_input": true,
-      "supports_pdf_input": true,
-      "supports_reasoning": true
-    },
-    "claude-sonnet-4-6-think": {
-      "context_window": 200000,
-      "max_output_tokens": 64000,
-      "supports_image_input": true,
-      "supports_pdf_input": true,
-      "supports_reasoning": true
-    },
-    "coding-glm-4.7": {
-      "context_window": 204800,
-      "max_output_tokens": 131072,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "coding-glm-4.7-free": {
-      "context_window": 204800,
-      "max_output_tokens": 131072,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "coding-glm-5-free": {
-      "context_window": 204800,
-      "max_output_tokens": 131072,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "coding-glm-5.1": {
-      "context_window": 200000,
-      "max_output_tokens": 128000,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "coding-minimax-m2.1-free": {
-      "context_window": 204800,
-      "max_output_tokens": 131072,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "deepseek-v3.2": {
-      "context_window": 131000,
-      "max_output_tokens": 64000,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "deepseek-v3.2-fast": {
-      "context_window": 128000,
-      "max_output_tokens": 128000,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "deepseek-v3.2-think": {
-      "context_window": 131000,
-      "max_output_tokens": 64000,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gemini-2.5-flash": {
-      "context_window": 1000000,
-      "max_output_tokens": 65000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "gemini-2.5-pro": {
-      "context_window": 2000000,
-      "max_output_tokens": 65000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gemini-3-pro-preview": {
-      "context_window": 1000000,
-      "max_output_tokens": 65000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gemini-3-pro-preview-search": {
-      "context_window": 1000000,
-      "max_output_tokens": 65000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "glm-4.6v": {
-      "context_window": 128000,
-      "max_output_tokens": 32768,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "glm-4.7": {
-      "context_window": 204800,
-      "max_output_tokens": 131072,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "glm-5": {
-      "context_window": 204800,
-      "max_output_tokens": 131072,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "glm-5.1": {
-      "context_window": 200000,
-      "max_output_tokens": 128000,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-4.1": {
-      "context_window": 1047576,
-      "max_output_tokens": 32768,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "gpt-4.1-mini": {
-      "context_window": 1047576,
-      "max_output_tokens": 32768,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "gpt-4.1-nano": {
-      "context_window": 1047576,
-      "max_output_tokens": 32768,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "gpt-4o": {
-      "context_window": 128000,
-      "max_output_tokens": 16384,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "gpt-5": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5-codex": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5-mini": {
-      "context_window": 200000,
-      "max_output_tokens": 64000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5-nano": {
-      "context_window": 128000,
-      "max_output_tokens": 16384,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "gpt-5-pro": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5.1": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5.1-codex": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5.1-codex-max": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5.1-codex-mini": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5.2": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5.2-codex": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5.4": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "gpt-5.4-mini": {
-      "context_window": 400000,
-      "max_output_tokens": 128000,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "kimi-k2.5": {
-      "context_window": 262144,
-      "max_output_tokens": 262144,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "minimax-m2.1": {
-      "context_window": 204800,
-      "max_output_tokens": 131072,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "minimax-m2.5": {
-      "context_window": 204800,
-      "max_output_tokens": 131072,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "o4-mini": {
-      "context_window": 200000,
-      "max_output_tokens": 65536,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "qwen3-235b-a22b-instruct-2507": {
-      "context_window": 262144,
-      "max_output_tokens": 262144,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "qwen3-235b-a22b-thinking-2507": {
-      "context_window": 262144,
-      "max_output_tokens": 262144,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    },
-    "qwen3-coder-480b-a35b-instruct": {
-      "context_window": 262144,
-      "max_output_tokens": 131000,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "qwen3-coder-next": {
-      "context_window": 262144,
-      "max_output_tokens": 65536,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "qwen3-max-2026-01-23": {
-      "context_window": 262144,
-      "max_output_tokens": 65536,
-      "supports_image_input": false,
-      "supports_pdf_input": false,
-      "supports_reasoning": false
-    },
-    "qwen3.5-plus": {
-      "context_window": 1000000,
-      "max_output_tokens": 65536,
-      "supports_image_input": true,
-      "supports_pdf_input": false,
-      "supports_reasoning": true
-    }
-  },
   "anthropic": {
     "claude-3-5-haiku-20241022": {
       "context_window": 200000,
@@ -891,6 +525,13 @@
       "supports_image_input": true,
       "supports_pdf_input": false,
       "supports_reasoning": true
+    },
+    "kimi-k2.6": {
+      "context_window": 262144,
+      "max_output_tokens": 262144,
+      "supports_image_input": true,
+      "supports_pdf_input": false,
+      "supports_reasoning": true
     }
   },
   "openai": {
@@ -1827,6 +1468,13 @@
       "supports_pdf_input": false,
       "supports_reasoning": true
     },
+    "moonshotai/kimi-k2.6": {
+      "context_window": 262144,
+      "max_output_tokens": 262144,
+      "supports_image_input": true,
+      "supports_pdf_input": false,
+      "supports_reasoning": true
+    },
     "nousresearch/hermes-3-llama-3.1-405b:free": {
       "context_window": 131072,
       "max_output_tokens": 131072,

{mycode_sdk-0.5.2 → mycode_sdk-0.5.3}/src/mycode/providers/anthropic_like.py

@@ -311,7 +311,7 @@ class AnthropicAdapter(AnthropicLikeAdapter):
     provider_id = "anthropic"
     label = "Anthropic"
     default_base_url = "https://api.anthropic.com"
-    env_api_key_names = ("ANTHROPIC_API_KEY", "ANTHROPIC_AUTH_TOKEN")
+    env_api_key_names = ("ANTHROPIC_API_KEY",)
     default_models = ("claude-sonnet-4-6", "claude-opus-4-7")
     supports_reasoning_effort = True
 
@@ -352,7 +352,7 @@ class AnthropicAdapter(AnthropicLikeAdapter):
 class MoonshotAIAdapter(AnthropicLikeAdapter):
     """Moonshot's Anthropic-compatible Messages endpoint.
 
-    kimi-k2.5 tool loops work through this endpoint. When thinking is enabled,
+    kimi-k2.6 tool loops work through this endpoint. When thinking is enabled,
     prior reasoning blocks must be replayed in the conversation history —
     Moonshot does not strip them on the server side.
     """
@@ -361,7 +361,7 @@ class MoonshotAIAdapter(AnthropicLikeAdapter):
     label = "Moonshot"
     default_base_url = "https://api.moonshot.ai/anthropic"
     env_api_key_names = ("MOONSHOT_API_KEY",)
-    default_models = ("kimi-k2.5",)
+    default_models = ("kimi-k2.6",)
     supports_reasoning_effort = True
 
     def thinking_config(self, request: ProviderRequest) -> dict[str, Any] | None:

mycode_sdk-0.5.2/README.md

@@ -1,84 +0,0 @@
-# mycode-sdk
-
-Lightweight Python SDK for building agents.
-
-## Install
-
-```bash
-uv add mycode-sdk
-# or
-pip install mycode-sdk
-```
-
-## Quick start
-
-```python
-import asyncio
-
-from mycode import Agent, bash_tool, read_tool
-
-
-async def main() -> None:
-    agent = Agent(
-        model="claude-sonnet-4-6",
-        api_key="YOUR_API_KEY",
-        tools=[read_tool, bash_tool],
-    )
-
-    async for event in agent.achat("Read pyproject.toml and tell me the project name."):
-        if event.type == "text":
-            print(event.data["delta"], end="", flush=True)
-
-
-asyncio.run(main())
-```
-
-`Agent(...)` infers the provider from the model id and defaults `cwd` to the current working directory. No tools are registered unless you pass `tools=[...]`, and nothing is persisted unless you pass `session_dir=`.
-
-For a synchronous call, use `run()` — it collects the stream into a `RunResult`:
-
-```python
-result = agent.run("Read pyproject.toml and tell me the project name.")
-print(result.text)
-```
-
-Call `achat` or `run` again on the same `Agent` to continue the conversation — history accumulates in `agent.messages`.
-
-To persist across processes, pass `session_dir` as the root directory; each `session_id` becomes a subdirectory. Reconstruct with the same `(session_dir, session_id)` to resume.
-
-## Built-in tools
-
-```python
-from mycode import read_tool, write_tool, edit_tool, bash_tool
-```
-
-Only `bash_tool` streams incremental output as `tool_output` events; the others return a single result.
-
-## Custom tools
-
-`@tool` wraps a sync or `async def` Python function as a `ToolSpec`. Parameter type hints become the JSON schema sent to the provider.
-
-Annotate the first parameter as `ToolContext` to have the context injected. Use `ctx.read / ctx.write / ctx.edit / ctx.bash` to invoke the built-ins, or `ctx.call(name, args)` for any registered tool by name.
-
-```python
-from mycode import Agent, ToolContext, read_tool, tool
-
-
-@tool
-def summarize_file(ctx: ToolContext, path: str) -> str:
-    """Return the first line of a text file."""
-
-    result = ctx.read(path)
-    return result.output.splitlines()[0] if result.output else ""
-
-
-agent = Agent(
-    model="claude-sonnet-4-6",
-    api_key="YOUR_API_KEY",
-    tools=[read_tool, summarize_file],
-)
-```
-
-A bare `str` return becomes the tool `output`; any other JSON-serializable value is dumped to JSON. For finer control, return a `ToolExecutionResult` to set `output`, `content` (multimodal blocks such as images), `metadata` (structured UI data), and `is_error` independently.
-
-See [docs/sdk.md](../docs/sdk.md) for the event stream, cancellation, session rules, and the full `Agent` / `@tool` reference.