bee-sdk 0.2.2__tar.gz → 0.3.0__tar.gz

{bee_sdk-0.2.2 → bee_sdk-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: bee-sdk
-Version: 0.2.2
+Version: 0.3.0
 Summary: Official Python client + MCP server for the Bee Intelligence Engine — domain-specialized LoRA-routed LLM by CUI Labs.
 Project-URL: Homepage, https://bee.cuilabs.io
 Project-URL: Documentation, https://bee.cuilabs.io/docs/sdks
@@ -41,7 +41,7 @@ Official Python client + MCP server for the **Bee Intelligence Engine** — a do
 
 <!-- mcp-name: io.github.cuilabs/bee -->
 
-SDK version: `0.2.2`.
+SDK version: `0.3.0` — adds per-call/`BEE_MODEL` model-tier selection (bee-cell … bee-swarm; access follows your key's plan) and MCP tool annotations.
 
 Includes a hosted **Model Context Protocol** server: `pip install bee-sdk` then
 run `bee-mcp` (stdio) to expose Bee's 11 domain tools to Claude Desktop, Cursor,

{bee_sdk-0.2.2 → bee_sdk-0.3.0}/README.md

@@ -4,7 +4,7 @@ Official Python client + MCP server for the **Bee Intelligence Engine** — a do
 
 <!-- mcp-name: io.github.cuilabs/bee -->
 
-SDK version: `0.2.2`.
+SDK version: `0.3.0` — adds per-call/`BEE_MODEL` model-tier selection (bee-cell … bee-swarm; access follows your key's plan) and MCP tool annotations.
 
 Includes a hosted **Model Context Protocol** server: `pip install bee-sdk` then
 run `bee-mcp` (stdio) to expose Bee's 11 domain tools to Claude Desktop, Cursor,

{bee_sdk-0.2.2 → bee_sdk-0.3.0}/bee_sdk/__init__.py

@@ -27,7 +27,7 @@ VS Code, …) this package ships a hosted MCP server — run ``bee-mcp``
 from .client import Bee, AsyncBee, BeeError, RateLimitError, BeeAPIError
 from .types import ChatMessage, ChatResponse, Domain, ModelTier
 
-__version__ = "0.2.2"
+__version__ = "0.3.0"
 __all__ = [
     "Bee",
     "AsyncBee",

{bee_sdk-0.2.2 → bee_sdk-0.3.0}/bee_sdk/client.py

@@ -36,6 +36,16 @@ DEFAULT_TIMEOUT = 60
 DEFAULT_RETRIES = 3
 RETRYABLE_STATUS = {429, 500, 502, 503, 504}
 
+# Model tier sent in the request body. Which tiers an API key may use is
+# decided by the key's plan at the gateway (403 model_access_denied when the
+# plan doesn't include it). bee-cell is the free tier, so it's the safe
+# default; override per-call or via BEE_MODEL.
+DEFAULT_MODEL = "bee-cell"
+
+
+def _resolve_model(model: Optional[str]) -> str:
+    return model or os.environ.get("BEE_MODEL") or DEFAULT_MODEL
+
 
 class BeeError(Exception):
     """Base error class for the SDK."""
@@ -69,7 +79,7 @@ class _BaseClient:
         h = {
             "Content-Type": "application/json",
             "Accept": "application/json",
-            "User-Agent": "bee-sdk/0.2.2",
+            "User-Agent": "bee-sdk/0.3.0",
         }
         if self.api_key:
             h["Authorization"] = f"Bearer {self.api_key}"
@@ -122,16 +132,21 @@ class Bee(_BaseClient):
         max_tokens: int = 512,
         temperature: float = 0.3,
         system: Optional[str] = None,
+        model: Optional[str] = None,
     ) -> str:
         """Single-turn chat. Returns the assistant text only.
 
-        For multi-turn or detailed response metadata, use `chat_messages`.
+        `model` picks the Bee tier (bee-cell … bee-swarm); which tiers your
+        key may use is governed by its plan. Defaults to BEE_MODEL env or
+        bee-cell. For multi-turn or detailed metadata, use `chat_messages`.
         """
         msgs: list[ChatMessage] = []
         if system:
             msgs.append(ChatMessage(role="system", content=system))
         msgs.append(ChatMessage(role="user", content=message))
-        return self.chat_messages(msgs, domain=domain, max_tokens=max_tokens, temperature=temperature).content
+        return self.chat_messages(
+            msgs, domain=domain, max_tokens=max_tokens, temperature=temperature, model=model
+        ).content
 
     def chat_messages(
         self,
@@ -139,6 +154,7 @@ class Bee(_BaseClient):
         domain: Domain = "general",
         max_tokens: int = 512,
         temperature: float = 0.3,
+        model: Optional[str] = None,
     ) -> ChatResponse:
         # Switch domain first (the Bee API has explicit /domain/switch).
         try:
@@ -148,7 +164,7 @@ class Bee(_BaseClient):
             # that domain's adapter, fall through and serve from current.
             pass
         body = {
-            "model": "bee-cell",
+            "model": _resolve_model(model),
             "messages": [{"role": m.role, "content": m.content} for m in messages],
             "max_tokens": max_tokens,
             "temperature": temperature,
@@ -174,6 +190,7 @@ class Bee(_BaseClient):
         max_tokens: int = 512,
         temperature: float = 0.3,
         system: Optional[str] = None,
+        model: Optional[str] = None,
     ) -> Iterator[str]:
         """Stream tokens as they're generated.
 
@@ -190,7 +207,7 @@ class Bee(_BaseClient):
             msgs.append({"role": "system", "content": system})
         msgs.append({"role": "user", "content": message})
         body = {
-            "model": "bee-cell",
+            "model": _resolve_model(model),
             "messages": msgs,
             "max_tokens": max_tokens,
             "temperature": temperature,
@@ -267,6 +284,7 @@ class AsyncBee(_BaseClient):
         max_tokens: int = 512,
         temperature: float = 0.3,
         system: Optional[str] = None,
+        model: Optional[str] = None,
     ) -> str:
         import httpx
         msgs: list[dict] = []
@@ -279,7 +297,7 @@ class AsyncBee(_BaseClient):
             except httpx.HTTPError:
                 pass
             r = await cl.post(f"{self.base_url}/chat/completions", json={
-                "model": "bee-cell",
+                "model": _resolve_model(model),
                 "messages": msgs,
                 "max_tokens": max_tokens,
                 "temperature": temperature,
@@ -295,6 +313,7 @@ class AsyncBee(_BaseClient):
         max_tokens: int = 512,
         temperature: float = 0.3,
         system: Optional[str] = None,
+        model: Optional[str] = None,
     ) -> AsyncIterator[str]:
         import httpx
         msgs: list[dict] = []
@@ -307,7 +326,7 @@ class AsyncBee(_BaseClient):
             except httpx.HTTPError:
                 pass
             async with cl.stream("POST", f"{self.base_url}/chat/completions", json={
-                "model": "bee-cell",
+                "model": _resolve_model(model),
                 "messages": msgs,
                 "max_tokens": max_tokens,
                 "temperature": temperature,

{bee_sdk-0.2.2 → bee_sdk-0.3.0}/bee_sdk/mcp.py

@@ -50,6 +50,18 @@ except Exception:  # not installed (running from source) — fall back
 # fallback when the client omits it.
 DEFAULT_PROTOCOL_VERSION = "2024-11-05"
 
+# Production model tiers — mirror of bee_sdk.types.ModelTier (customer-
+# selectable ladder; enclave/ignite are not public-selectable). Which tiers a
+# key may use is decided by its plan at the gateway; bee-cell is the free tier.
+MODELS = [
+    "bee-cell",
+    "bee-brood",
+    "bee-comb",
+    "bee-buzz",
+    "bee-hive",
+    "bee-swarm",
+]
+
 # Authoritative domain list — mirror of bee_sdk.types.Domain (itself a
 # mirror of the private core's bee/domains.py TIER_1_DOMAINS). Kept inline
 # so the tool enum doesn't depend on the private core being importable.
@@ -72,12 +84,13 @@ TOOLS = [
     {
         "name": "bee_chat",
         "description": (
-            "Ask Bee a question. Bee is a domain-specialized small LLM with "
-            "per-domain LoRA adapters. Specialised in: programming, AI/ML, "
-            "cybersecurity, quantum computing, fintech, blockchain, cloud "
-            "infrastructure, research methodology, and business operations. "
-            "Use Bee for technical depth on these domains; Bee is honest about "
-            "uncertainty rather than fabricating."
+            "Ask Bee a question. Bee — The Progressive Intelligence Engine — "
+            "is a tiered LLM platform with per-domain LoRA specialisation: "
+            "programming, AI/ML, cybersecurity, quantum computing, fintech, "
+            "blockchain, cloud infrastructure, research methodology, and "
+            "business operations. Optionally pick a model tier (bee-cell free "
+            "… bee-swarm premium); tier access follows the API key's plan. "
+            "Bee is honest about uncertainty rather than fabricating."
         ),
         "inputSchema": {
             "type": "object",
@@ -89,6 +102,15 @@ TOOLS = [
                     "enum": DOMAINS,
                     "default": "general",
                 },
+                "model": {
+                    "type": "string",
+                    "description": (
+                        "Bee model tier. Access is governed by the API key's plan "
+                        "(403 if the plan doesn't include it). Default: BEE_MODEL "
+                        "env or bee-cell (free tier)."
+                    ),
+                    "enum": MODELS,
+                },
                 "max_tokens": {"type": "integer", "description": "Max response tokens", "default": 512},
             },
             "required": ["message"],
@@ -238,6 +260,13 @@ TOOLS = [
     },
 ]
 
+# MCP ToolAnnotations (behaviour hints; required by directory reviewers).
+# Every Bee tool is identical on these axes: read-only (it forwards a prompt
+# to the hosted gateway and modifies nothing on the user's machine) and
+# open-world (it talks to an external service).
+for _tool in TOOLS:
+    _tool["annotations"] = {"readOnlyHint": True, "openWorldHint": True}
+
 RESOURCES = [
     {
         "uri": "bee://status",
@@ -266,6 +295,8 @@ def handle_tool_call(client: Bee, name: str, arguments: dict[str, Any]) -> str:
             message=arguments["message"],
             domain=domain,
             max_tokens=arguments.get("max_tokens", 512),
+            # None falls through to BEE_MODEL env / bee-cell in the client.
+            model=arguments.get("model"),
             system=(
                 f"You are Bee, a domain-specialized AI expert in {domain}. "
                 "Be precise and thorough. Admit uncertainty rather than fabricate."

bee_sdk-0.3.0/mcpb/pyproject.toml

@@ -0,0 +1,8 @@
+[project]
+name = "bee-mcpb"
+version = "0.3.0"
+description = "Bee MCP bundle launcher — runs the hosted bee-mcp server from bee-sdk"
+requires-python = ">=3.10"
+dependencies = [
+    "bee-sdk>=0.3.0",
+]

{bee_sdk-0.2.2 → bee_sdk-0.3.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "bee-sdk"
-version = "0.2.2"
+version = "0.3.0"
 description = "Official Python client + MCP server for the Bee Intelligence Engine — domain-specialized LoRA-routed LLM by CUI Labs."
 readme = "README.md"
 requires-python = ">=3.10"