LLMs & Chat Models

from standins import RuleModel
from langchain.messages import SystemMessage, HumanMessage, AIMessage

# In Colab: llm = ChatGroq(model="openai/gpt-oss-20b")
llm = RuleModel(rules={
    "daily life": "Google Maps uses AI to predict traffic and suggest the fastest route to office.",
    "gravity": "Gravity is the Earth's hug that never lets go! Drop your phone and it says hello to the floor.",
    "taste was bad": "NEGATIVE",
    "mumbai rains": "Grey clouds over Marine Drive,\nvada pav and cutting chai,\nlocal trains run, umbrellas fly,\nMumbai smiles at the monsoon sky.",
    "what is ai": "AI is making computers do tasks that need human intelligence.",
    "what is ml": "ML is a part of AI where computers learn patterns from data.",
    "deep learning": "Deep learning uses many-layered neural networks to learn from huge data.",
})

# 1. invoke
response = llm.invoke("Give me a short example of AI use in daily life?")
print(type(response).__name__, "->", response.content)
print("usage:", response.usage_metadata)

# 2. messages with roles
messages = [SystemMessage("You are a funny teacher. Answer in 2 lines."), HumanMessage("What is gravity?")]
print("\nteacher:", llm.invoke(messages).content)

# 3. few-shot with AIMessage examples
messages = [
    SystemMessage("Classify the review as POSITIVE or NEGATIVE. One word only."),
    HumanMessage("The food was amazing"), AIMessage("POSITIVE"),
    HumanMessage("Delivery was very late"), AIMessage("NEGATIVE"),
    HumanMessage("Packaging was good but taste was bad"),
]
print("few-shot:", llm.invoke(messages).content)

# 4. stream word by word
print("\nstream: ", end="")
chunks = 0
for chunk in llm.stream("Write a 4-line poem on Mumbai rains"):
    print(chunk.content, end="|", flush=True); chunks += 1
print(f"\n({chunks} chunks)")

# 5. batch many questions
questions = ["What is AI?", "What is ML?", "What is Deep Learning?"]
for q, r in zip(questions, llm.batch(questions)):
    print(q, "->", r.content)

profile = llm.profile or {}
print("Context window :", profile.get("max_input_tokens"))
print("Max output     :", profile.get("max_output_tokens"))
print("Tool calling   :", profile.get("tool_calling"))
print("Structured out :", profile.get("structured_output"))
print("Image input    :", profile.get("image_inputs"))

from langchain.chat_models import init_chat_model

llm = init_chat_model("groq:openai/gpt-oss-20b", temperature=0)      # Groq
llm = init_chat_model("google_genai:gemini-2.5-flash")                # pip install langchain-google-genai
llm = init_chat_model("openai:gpt-4o-mini")                           # pip install langchain-openai
llm = init_chat_model("ollama:llama3.1")                              # free, runs on your laptop

print(llm.invoke("Capital of Maharashtra?").content)                   # same line for all