LiteLLM Lazy loading

docs/docs/communityhub/release_notes.md

@@ -5,9 +5,6 @@ This document provides a summary of new features, improvements, and bug fixes in
 
 ## jaclang 0.8.10 / jac-cloud 0.2.10 / byllm 0.4.5 (Unreleased)
 
-- **byLLM Lazy Loading**: Refactored byLLM to support lazy loading by moving all exports to `byllm.lib` module. Users should now import from `byllm.lib` in Python (e.g., `from byllm.lib import Model, by`) and use `import from byllm.lib { Model }` in Jac code. This improves startup performance and reduces unnecessary module loading.
-- **NonGPT Fallback for byLLM**: Implemented automatic fallback when byLLM is not installed. When code attempts to import `byllm`, the system will provide mock implementations that return random using the `NonGPT.random_value_for_type()` utility.
-
 ## jaclang 0.8.9 / jac-cloud 0.2.9 / byllm 0.4.4 (Latest Release)
 
 - **Typed Context Blocks (OSP)**: Fully implemented typed context blocks (`-> NodeType { }` and `-> WalkerType { }`) for Object-Spatial Programming, enabling conditional code execution based on runtime types.

docs/docs/jac_book/chapter_5.md

@@ -82,7 +82,7 @@ pip install byllm
 Next we replace the OpenAI import with that of the byLLM package
 
 ```jac
-import from byllm.lib { Model }
+import from byllm { Model }
 glob llm = Model(model_name="gpt-4.1-mini");
 ```
 <br />
@@ -94,7 +94,7 @@ def write_poetry(topic: str) -> str by llm();
 Finally, lets put it all together and run the Jac code:
 ```jac
 # mt_poem.jac - Simple AI integration
-import from byllm.lib { Model }
+import from byllm { Model }
 
 glob llm = Model(model_name="gpt-4.1-mini");
 
@@ -137,7 +137,7 @@ Next we'll make use of MLTLLM's `Image` function to handle image inputs. This fu
 
 ```jac
 # image_captioning.jac - Simple Image Captioning Tool
-import from byllm.lib { Model, Image }
+import from byllm { Model, Image }
 
 glob llm = Model(model_name="gpt-4o-mini");
 
@@ -169,7 +169,7 @@ byLLM supports various AI models through the unified `Model` interface. For exam
 
 ```jac
 # basic_setup.jac
-import from byllm.lib { Model, Image }
+import from byllm { Model, Image }
 
 # Configure different models
 glob text_model = Model(model_name="gpt-4o");
@@ -182,7 +182,7 @@ glob gemini_model = Model(model_name="gemini-2.0-flash");
 The `Model` class allows you to configure various parameters for your AI model, such as temperature, max tokens, and more. Here's an example of how to set up a model with custom parameters:
 
 ```jac
-import from byllm.lib { Model, Image }
+import from byllm { Model, Image }
 
 # Configure model with custom parameters
 glob creative_model = Model(
@@ -210,7 +210,7 @@ Below is a breakdown of the parameters you can configure when creating a `Model`
 Here we have a simple example of how to use the `Model` class to create a model instance with custom parameters:
 ```jac
 # model_config.jac
-import from byllm.lib { Model, Image }
+import from byllm { Model, Image }
 
 # Configure model with custom parameters
 glob creative_model = Model(
@@ -252,7 +252,7 @@ Let's progressively build an image captioning tool that demonstrates byLLM's cap
 
 ```jac
 # image_captioner.jac
-import from byllm.lib { Model, Image }
+import from byllm { Model, Image }
 
 glob vision_llm = Model(model_name="gpt-4o-mini");
 
@@ -307,7 +307,7 @@ the stylish outfit of the dog contribute to a fun and lighthearted atmosphere.
 
 ```jac
 # enhanced_captioner.jac
-import from byllm.lib { Model, Image }
+import from byllm { Model, Image }
 
 glob vision_llm = Model(model_name="gpt-4.1-mini");
 
@@ -363,7 +363,7 @@ AI applications require robust error handling and testing strategies.
 
 ```jac
 # robust_ai.jac
-import from byllm.lib { Model, Image }
+import from byllm { Model, Image }
 
 glob reliable_llm = Model(model_name="gpt-4o", max_tries=3);
 

docs/docs/jac_book/chapter_9.md

@@ -197,7 +197,7 @@ Using the byLLM plugin, we can define a function that sends the agent's state to
 
 Here’s the code for our mood function:
 ```jac
-import from byllm.lib { Model }
+import from byllm { Model }
 
 # Configure the LLM
 glob npc_model = Model(model_name="gpt-4.1-mini");
@@ -242,7 +242,7 @@ The `NPCWalker` first inherits the behavior of `StateAgent` (which collects cont
 
 Finally, we can compose everything in a single entry point:
 ```jac
-import from byllm.lib { Model }
+import from byllm { Model }
 
 # Configure different models
 glob npc_model = Model(model_name="gpt-4.1-mini");

docs/docs/learn/beginners_guide_to_jac.md

@@ -2678,7 +2678,7 @@ Now that you have the foundation, here are advanced Jac features to explore:
    <div class="code-block">
 
    ```jac
-   import from byllm.lib { Model }
+   import from byllm { Model }
 
    glob llm = Model(model="gpt-4");
 

docs/docs/learn/examples/agentic_ai/aider-genius-lite/genius_lite.jac

@@ -1,5 +1,5 @@
 
-import from byllm.lib { Model }
+import from byllm { Model }
 import from pathlib { Path }
 
 glob llm = Model(model_name="gpt-4o-mini");

docs/docs/learn/examples/littleX/src/byllm_example.jac

@@ -1,4 +1,4 @@
-import from byllm.lib {Model}
+import from byllm {Model}
 
 glob llm = Model(host="http://127.0.0.1:11434", model_name="ollama/llama3.2:1b");
 

docs/docs/learn/examples/littleX/src/example_game.jac

@@ -1,4 +1,4 @@
-import from byllm.lib {Model}
+import from byllm {Model}
 
 glob llm = Model(model_name="gpt-4o");
 

docs/docs/learn/examples/mtp_examples/fantasy_trading_game.md

@@ -59,7 +59,7 @@ obj Chat {
 Configure the LLM for AI operations:
 
 ```jac
-import from byllm.lib {Model}
+import from byllm {Model}
 
 glob llm = Model(model_name="gpt-4o");
 ```

docs/docs/learn/examples/mtp_examples/rpg_game.md

@@ -114,7 +114,7 @@ We’ll connect to an LLM (GPT-4o here) and define AI-powered methods for genera
 At the top of `level_manager.jac`, import the model:
 
 ```jac
-import from byllm.lib { Model }
+import from byllm { Model }
 
 glob llm = Model(model_name="gpt-4o", verbose=True);
 ```

docs/docs/learn/jac-byllm/litellm_proxy.md

@@ -11,7 +11,7 @@ Reference: [https://docs.litellm.ai/docs/proxy/deploy](https://docs.litellm.ai/d
 Once The proxy server is setted up and running, you can connect to it by simply passing the URL of the proxy server to the byLLM model with the parameter `proxy_url`:
 
 ```python
-from byllm.lib import Model
+from byllm import Model
 
 llm = Model(
     model_name="gpt-4o",                # The model name to be used

docs/docs/learn/jac-byllm/multimodality.md

@@ -13,7 +13,7 @@ pip install byllm[video]
 byLLM supports image inputs through the `Image` format. Images can be provided as input to byLLM functions or methods:
 
 ```jac
-import from byllm.lib { Model, Image }
+import from byllm { Model, Image }
 
 glob llm = Model(model_name="gpt-4o");
 
@@ -57,7 +57,7 @@ In this example, an image of a person is provided as input to the `get_person_in
 byLLM supports video inputs through the `Video` format. Videos can be provided as input to byLLM functions or methods:
 
 ```jac
-import from byllm.lib { Model, Video }
+import from byllm { Model, Video }
 
 glob llm = Model(model_name="gpt-4o");
 

docs/docs/learn/jac-byllm/python_integration.md

@@ -15,7 +15,7 @@ byLLM functionality is accessed by importing the `byllm` module and using the `b
 ```python linenums="1"
 import jaclang
 from dataclasses import dataclass
-from byllm.lib import Model, Image, by
+from byllm import Model, Image, by
 
 llm = Model(model_name="gpt-4o")
 
@@ -43,7 +43,7 @@ print(f"Name: {person.full_name}, Description: {person.description}, Year of Bir
 In Jaclang, hyper-parameters are set by passing them to the LLM model:
 
 ```jac linenums="1"
-import from byllm.lib { Model }
+import from byllm { Model }
 
 glob llm = Model(model_name="gpt-4o")
 
@@ -56,7 +56,7 @@ In Python, hyper-parameters are passed as follows:
 
 ```python linenums="1"
 import jaclang
-from byllm.lib import Model, by
+from byllm import Model, by
 
 llm = Model(model_name="gpt-4o")
 
@@ -70,7 +70,7 @@ Python functions can be used as tools in byLLM. Functions defined in Python are
 
 ```python linenums="1"
 import jaclang
-from byllm.lib import Model
+from byllm import Model
 llm = Model(model_name="gpt-4o")
 
 
@@ -98,7 +98,6 @@ Using `sem` functionality in python is a bit diferent as the attachment is done
 
 ```python
 from jaclang import JacMachineInterface as Jac
-from byllm.lib import Model, by
 
 @Jac.sem('<Person Semstring>', {
     'name' : '<name semstring>',

docs/docs/learn/jac-byllm/quickstart.md

@@ -44,7 +44,7 @@ The `by` keyword abstraction enables functions to process inputs of any type and
 #### Step 1: Configure LLM Model
 
 ```jac linenums="1"
-import from byllm.lib {Model}
+import from byllm {Model}
 
 glob llm = Model(model_name="gemini/gemini-2.0-flash");
 ```
@@ -77,7 +77,7 @@ As byLLM is a python package, it can be natively used in jac. The following code
 
 ```python linenums="1"
 import jaclang
-from byllm.lib import Model, by
+from byllm import Model, by
 from enum import Enum
 
 llm = Model(model_name="gemini/gemini-2.0-flash")

docs/docs/learn/jac-byllm/usage.md

@@ -8,31 +8,31 @@ byLLM uses [LiteLLM](https://docs.litellm.ai/docs) to provide integration with a
 
 === "OpenAI"
     ```jac linenums="1"
-    import from byllm.lib {Model}
+    import from byllm {Model}
 
     glob llm = Model(model_name = "gpt-4o")
     ```
 === "Gemini"
     ```jac linenums="1"
-    import from byllm.lib {Model}
+    import from byllm {Model}
 
     glob llm = Model(model_name = "gemini/gemini-2.0-flash")
     ```
 === "Anthropic"
     ```jac linenums="1"
-    import from byllm.lib {Model}
+    import from byllm {Model}
 
     glob llm = Model(model_name = "claude-3-5-sonnet-20240620")
     ```
 === "Ollama"
     ```jac linenums="1"
-    import from byllm.lib {Model}
+    import from byllm {Model}
 
     glob llm = Model(model_name = "ollama/llama3:70b")
     ```
 === "HuggingFace Models"
     ```jac linenums="1"
-    import from byllm.lib {Model}
+    import from byllm {Model}
 
     glob llm = Model(model_name = "huggingface/meta-llama/Llama-3.3-70B-Instruct")
     ```
@@ -270,7 +270,7 @@ In this example:
 The ReAct (Reasoning and Acting) method enables agentic behavior by allowing functions to reason about problems and use external tools. Functions can be made agentic by adding the `by llm(tools=[...])` declaration.
 
 ```jac linenums="1"
-import from byllm.lib { Model }
+import from byllm { Model }
 import from datetime { datetime }
 
 glob llm = Model(model_name="gpt-4o");
@@ -304,7 +304,7 @@ The streaming feature enables real-time token reception from LLM functions, usef
 Set `stream=True` in the invoke parameters to enable streaming:
 
 ```jac linenums="1"
-import from byllm.lib { Model }
+import from byllm { Model }
 
 glob llm = Model(model_name="gpt-4o-mini");
 

docs/docs/learn/jac-byllm/with_llm.md

@@ -43,7 +43,7 @@ pip install byllm
 Consider building an application that translates english to other languages using an LLM. This can be simply built as follows:
 === "Jac"
     ```jac linenums="1"
-    import from byllm.lib { Model }
+    import from byllm { Model }
 
     glob llm = Model(model_name="gpt-4o");
 
@@ -56,7 +56,7 @@ Consider building an application that translates english to other languages usin
     ```
 === "python"
     ```python linenums="1"
-    from byllm.lib import Model, by
+    from byllm import Model, by
 
     llm = Model(model_name="gpt-4o")
 
@@ -75,7 +75,7 @@ Consider a program that detects the personality type of a historical figure from
 
 === "Jac"
     ```jac linenums="1"
-    import from byllm.lib { Model }
+    import from byllm { Model }
     glob llm = Model(model_name="gemini/gemini-2.0-flash");
 
     enum Personality {
@@ -94,7 +94,7 @@ Consider a program that detects the personality type of a historical figure from
     ```
 === "Python"
     ```python linenums="1"
-    from byllm.lib import Model, by
+    from byllm import Model, by
     from enum import Enum
     llm =  Model(model_name="gemini/gemini-2.0-flash")
 
@@ -119,7 +119,7 @@ Even if we are elimination prompt engineering entierly, we allow specific ways t
 
 === "Jac"
     ```jac linenums="1"
-    import from byllm.lib { Model }
+    import from byllm { Model }
     glob llm = Model(model_name="gemini/gemini-2.0-flash");
 
     """Represents the personal record of a person"""
@@ -140,7 +140,7 @@ Even if we are elimination prompt engineering entierly, we allow specific ways t
     ```python linenums="1"
     from jaclang import JacMachineInterface as Jac
     from dataclasses import dataclass
-    from byllm.lib import Model, by
+    from byllm import Model, by
     llm =  Model(model_name="gemini/gemini-2.0-flash")
 
     @Jac.sem('', {  'name': 'Full name of the person',

docs/docs/learn/tour.md

@@ -35,7 +35,7 @@ This snippet natively imports Python packages `math` and `random` and runs ident
 Jac provides novel constructs for integrating LLMs into code. A function body can simply be replaced with a call to an LLM, removing the need for prompt engineering or extensive use of new libraries.
 
 ```jac
-import from byllm.lib { Model }
+import from byllm { Model }
 glob llm = Model(model_name="gpt-4o");
 
 enum Personality {

jac-byllm/README.md

@@ -26,7 +26,7 @@ pip install byllm
 Consider building an application that translates english to other languages using an LLM. This can be simply built as follows:
 
 ```python
-import from byllm.lib { Model }
+import from byllm { Model }
 
 glob llm = Model(model_name="gpt-4o");
 
@@ -45,7 +45,7 @@ This simple piece of code replaces traditional prompt engineering without introd
 Consider a program that detects the personality type of a historical figure from their name. This can eb built in a way that LLM picks from an enum and the output strictly adhere this type.
 
 ```python
-import from byllm.lib { Model }
+import from byllm { Model }
 glob llm = Model(model_name="gemini/gemini-2.0-flash");
 
 enum Personality {

jac-byllm/byllm/__init__.py

@@ -1 +1,10 @@
-"""byLLM Package - Lazy Loading."""
+"""byLLM Package."""
+
+from byllm.llm import Model
+from byllm.mtir import MTIR
+from byllm.plugin import JacMachine
+from byllm.types import Image, MockToolCall, Video
+
+by = JacMachine.by
+
+__all__ = ["by", "Image", "MockToolCall", "Model", "MTIR", "Video"]