Ch 2 — Your First LLM Chain
Prompt in, response out. The simplest possible LangChain call.
High Level
-
Click play or press Space to begin...
A
What Is a Chain?
The simplest unit of work in LangChain
1
input
Input
A question or
instruction (dict)
goes into
link
Chain
Prompt | Model
| Parser
produces
output
Output
A string, object,
or structured data
2
arrow_downward The three pieces of every chain
B
The Three Pieces
Prompt, Model, Parser
edit_note
Prompt
Template with
variables like {topic}
formats
smart_toy
Chat Model
GPT-4o, Claude,
Gemini, Llama...
returns
output
Output Parser
Extracts the text
or structured data
3
arrow_downward Step 1: Build the prompt template
C
The Prompt Template
Turns user input into formatted messages
person
System Message
"You are a
helpful assistant"
+
chat
Human Message
"Explain {topic}
in one sentence"
=
format_list_bulleted
Message List
Ready to send
to the model
4
arrow_downward Step 2: The model generates a response
D
The Chat Model
Sends messages to the LLM API and gets a response
format_list_bulleted
Messages In
[SystemMessage,
HumanMessage]
API call
cloud
LLM API
OpenAI, Anthropic,
Google, local...
returns
psychology
AIMessage
content + metadata
(tokens, model, etc.)
5
arrow_downward Step 3: Parse the output
E
The Output Parser
Extracts usable data from the AIMessage
psychology
AIMessage
Raw response
object from LLM
extract
text_fields
StrOutputParser
Pulls out the
.content string
or
6
data_object
JsonOutputParser
Parses JSON into
a Python dict
7
arrow_downward Putting it all together with the | operator
F
The Complete Chain
Three pieces composed into one callable
edit_note
prompt
ChatPrompt
Template
|
smart_toy
model
ChatOpenAI
(gpt-4o)
|
output
parser
StrOutput
Parser
.invoke()
8
check_circle
Result
"Quantum computing
uses qubits..."