Context formatting and injection in LangChain - Step-by-Step Execution

Concept Flow - Context formatting and injection

Start with raw context data

↓

Format context into template

↓

Inject formatted context into prompt

↓

Send prompt to language model

↓

Receive and process model output

This flow shows how raw context is first formatted into a template, then injected into a prompt before sending to the language model.

Execution Sample

LangChain

from langchain.prompts import PromptTemplate

context = "The sky is blue."
template = "Context: {context}\nQuestion: What color is the sky?"
prompt = PromptTemplate(input_variables=["context"], template=template)
formatted_prompt = prompt.format(context=context)

This code formats a context string into a prompt template by injecting the context into the placeholder.

Execution Table

Step	Action	Input	Output	Notes
1	Receive raw context	"The sky is blue."	"The sky is blue."	Raw context string stored
2	Define template	"Context: {context}\nQuestion: What color is the sky?"	Template with placeholder	Template ready for formatting
3	Create PromptTemplate	input_variables=["context"], template=template	PromptTemplate object	Setup for formatting
4	Format prompt	context="The sky is blue."	"Context: The sky is blue.\nQuestion: What color is the sky?"	Context injected into template
5	Send prompt to model	Formatted prompt string	Model output (e.g. "The sky is blue.")	Model processes prompt
6	Process output	Model output	Final answer	Use output as needed
7	End			Execution complete

💡 All steps complete; prompt formatted and sent to model, output processed.

Variable Tracker

Variable	Start	After Step 1	After Step 4	Final
context	None	"The sky is blue."	"The sky is blue."	"The sky is blue."
template	None	None	"Context: {context}\nQuestion: What color is the sky?"	"Context: {context}\nQuestion: What color is the sky?"
formatted_prompt	None	None	"Context: The sky is blue.\nQuestion: What color is the sky?"	"Context: The sky is blue.\nQuestion: What color is the sky?"

Key Moments - 2 Insights

Why do we need to format the context before sending it to the model?

Because the model expects a complete prompt string. Formatting injects the context into the template to create that prompt, as shown in step 4 of the execution_table.

What happens if the context variable is missing when formatting?

The formatting will fail or produce an incomplete prompt because the placeholder {context} has no value, as seen in step 4 where context is required.

Visual Quiz - 3 Questions

Look at the execution_table, what is the output after step 4?

A "Context: The sky is blue.\nQuestion: What color is the sky?"

B "The sky is blue."

C PromptTemplate object

D Model output

💡 Hint

Check the 'Output' column for step 4 in the execution_table.

At which step is the context injected into the template?

A Step 2

B Step 4

C Step 3

D Step 5

💡 Hint

Look for the step where formatting happens in the execution_table.

If the context changes to "Grass is green.", which variable changes in variable_tracker?

A template

B formatted_prompt

C Both B and D

D context

💡 Hint

See how context and formatted_prompt depend on the context variable in variable_tracker.

Concept Snapshot

Context formatting and injection:
- Define a template with placeholders like {context}
- Use PromptTemplate to prepare formatting
- Inject actual context into template with format()
- Send formatted prompt to language model
- Receive and use model output
Key: context must be provided to fill placeholders

Full Transcript

This visual execution shows how context formatting and injection works in Langchain. First, raw context data is received. Then a template string with placeholders is defined. A PromptTemplate object is created with the template and input variables. Next, the context is injected into the template using the format method, producing a complete prompt string. This prompt is sent to the language model, which returns an output. Finally, the output is processed for use. Variables like context, template, and formatted_prompt change values as the steps progress. Key points include the necessity of formatting before sending to the model and ensuring all placeholders have values to avoid errors.