Context Engineering by now. This text will cowl the important thing concepts behind creating LLM functions utilizing Context Engineering ideas, visually clarify these workflows, and share code snippets that apply these ideas virtually.
Don’t fear about copy-pasting the code from this text into your editor. On the finish of this text, I’ll share the GitHub hyperlink to the open-source code repository and a hyperlink to my 1-hour 20-minute YouTube course that explains the ideas offered right here in better element.
Until in any other case talked about, all pictures used on this article are produced by the writer and are free to make use of.
Let’s start!
What’s Context Engineering?
There’s a vital hole between writing easy prompts and constructing production-ready functions. Context Engineering is an umbrella time period that refers back to the delicate artwork and science of becoming data into the context window of an LLM as it really works on a job.
The precise scope of the place the definition of Context Engineering begins and ends is debatable, however in keeping with this tweet from Andrej Karpathy, we are able to determine the next key factors:
- It’s not simply atomic immediate engineering, the place you ask one query to the LLM and get a response
- It’s a holistic method that breaks up a bigger downside into a number of subproblems
- These subproblems may be solved by a number of LLMs (or brokers) in isolation. Every agent is supplied with the suitable context to hold out its job
- Every agent may be of acceptable functionality and dimension relying on the complexity of the duty.
- Intermediate steps that every agent can take to finish the duty – the context is not simply data we enter – it additionally contains intermediate tokens that the LLMs see throughout technology (eg. reasoning steps, software outcomes, and so on)
- The brokers are related with management flows, and we orchestrate precisely how data flows by means of our system
- The knowledge out there to the brokers can come from a number of sources – exterior databases with Retrieval-Augmented Technology (RAG), software calls (like net search), reminiscence techniques, or traditional few-shot examples.
- Brokers can take actions whereas producing responses. Every motion the agent can take needs to be well-defined so the LLM can work together with it by means of reasoning and appearing.
- Moreover, techniques should be evaluated with metrics and maintained with observability. Monitoring token utilization, latency, and price to output high quality is a key consideration.
Essential: How this text is structured
All through this text, I might be referring to the factors above whereas offering examples of how they’re utilized in constructing actual functions. At any time when I achieve this, I’ll use a block quote like this:
It’s a holistic method that breaks up a bigger downside into a number of subproblems
Once you see a quote on this format, the instance that follows will apply the quoted idea programmatically.
However earlier than that, we should ask ourselves one query…
Why not go the whole lot into the LLM?
Analysis has proven that cramming each piece of data into the context of an LLM is much from supreme. Regardless that many frontier fashions do declare to help “long-context” home windows, they nonetheless endure from points like context poisoning or context rot.
(Supply: Chroma)
An excessive amount of pointless data in an LLM’s context can pollute the mannequin’s understanding, result in hallucinations, and lead to poor efficiency.
That is why merely having a big context window isn’t sufficient. We want systematic approaches to context engineering.
Why DSPY
For this tutorial, I’ve chosen the DSPy framework. I’ll clarify the reasoning for this alternative shortly, however let me guarantee you that the ideas offered right here apply to virtually any prompting framework, together with writing prompts in pure English.
DSPy is a declarative framework for constructing modular AI software program. They’ve neatly separated the 2 key elements of any LLM job —
(a) the enter and output contracts handed right into a module,
and (b) the logic that governs how data flows.
Let’s see an instance!
Think about we need to use an LLM to put in writing a joke. Particularly, we would like it to generate a setup, a punchline, and the total supply in a comic’s voice.
Oh, and we additionally need the output in JSON format in order that we are able to post-process particular person fields of the dictionary after technology. For instance, maybe we need to print the punchline on a T-shirt (assume somebody has already written a handy perform for that).
system_prompt = """
You're a comic who tells jokes, you might be all the time humorous.
Generate the setup, punchline, and full supply within the comic's voice.
Output within the following JSON format:
{
"setup": ,
"punchline": ,
"supply":
}
Your response needs to be parsable withou errors in Python utilizing json.masses().
"""
consumer = openai.Consumer()
response = consumer.chat.completions.create(
mannequin="gpt-4o-mini",
temperature = 1,
messages=[
{"role": "system", "content": system_prompt,
{"role": "user", "content": "Write a joke about AI"}
]
)
joke = json.masses(response.decisions[0].message.content material) # Hope for the perfect
print_on_a_tshirt(joke["punchline"]) Discover how we post-process the LLM’s response to extract the dictionary? What if one thing “dangerous” occurred, just like the LLM failing to generate the response within the desired format? Our complete code would fail and there might be no printing on any T-shirts!
The above code can be fairly troublesome to increase. For instance, if we needed the LLM to do chain of thought reasoning earlier than producing the reply, we would want to put in writing further logic to parse that reasoning textual content accurately.
Moreover, it may be troublesome to have a look at plain English prompts like these and perceive what the inputs and outputs of those techniques are. DSPy solves all the above. Let’s write the above instance utilizing DSPy.
class JokeGenerator(dspy.Signature):
"""You are a comic who tells jokes. You are all the time humorous."""
question: str = dspy.InputField()
setup: str = dspy.OutputField()
punchline: str = dspy.OutputField()
supply: str = dspy.OutputField()
joke_gen = dspy.Predict(JokeGenerator)
joke_gen.set_lm(lm=dspy.LM("openai/gpt-4.1-mini", temperature=1))
outcome = joke_gen(question="Write a joke about AI")
print(outcome)
print_on_a_tshirt(outcome.punchline)
This method provides you structured, predictable outputs that you may work with programmatically, eliminating the necessity for regex parsing or error-prone string manipulation.
Dspy Signatures explicitly makes you outline what the inputs to the system are (“question” within the above instance), and the outputs to the system (setup, punchline, and supply) in addition to their data-types. It additionally tells the LLM the order during which you need them to be generated.
The dspy.Predict factor is an instance of a DSPy Module. With modules, you outline how the LLM converts from inputs to outputs. dspy.Predict is essentially the most fundamental one – you may go the question to it, as in joke_gen(question="Write a joke about AI") and it’ll create a fundamental immediate to ship to the LLM. Internally, DSPy simply creates a immediate as you may see under.
As soon as the LLM responds, DSPy will create Pydantic BaseModel objects that carry out automated schema validation and ship again the output. If errors happen throughout this validation course of, DSPy routinely makes an attempt to repair them by re-prompting the LLM—thereby considerably decreasing the danger of a program crash.
One other widespread theme in context engineering is Chain of Thought. Right here, we would like the LLM to generate reasoning textual content earlier than offering its last reply. This permits the LLM’s context to be populated with its self-generated reasoning earlier than it generates the ultimate output tokens.
To try this, you may merely substitute dspy.Predict with dspy.ChainOfThought within the instance above. The remainder of the code stays the identical. Now you may see that the LLM generates reasoning earlier than the outlined output fields.
Multi-Step Interactions and Agentic Workflows
One of the best a part of DSPy’s method is the way it decouples system dependencies (Signatures) from management flows (Modules), which makes writing code for multi-step interactions trivial (and enjoyable!). On this part, let’s see how we are able to construct some easy agentic flows.
Sequential Processing
Let’s remind ourselves about one of many key elements of Context Engineering.
It’s a holistic method that breaks up a bigger downside into a number of subproblems
Let’s proceed with our joke technology instance. We are able to simply separate out two subproblems from it. Producing the thought is one, making a joke is one other.
Let’s have two brokers then — the primary Agent generates a joke concept (setup and punchline) from a question. A second agent then generates the joke from this concept.
Every agent may be of acceptable functionality and dimension relying on the complexity of the duty
We’re additionally operating the primary agent with gpt-4.1-mini and the second agent with the extra highly effective gpt-4.1.
Discover how we wrote our personal dspy.Module known as JokeGenerator. Right here we use two separate dspy modules – the query_to_idea and the idea_to_joke to transform our authentic question to a JokeIdea and subsequently right into a joke (as pictured above).
class JokeIdea(BaseModel):
setup: str
contradiction: str
punchline: str
class QueryToIdea(dspy.Signature):
"""Generate a joke concept with setup, contradiction, and punchline."""
question = dspy.InputField()
joke_idea: JokeIdea = dspy.OutputField()
class IdeaToJoke(dspy.Signature):
"""Convert a joke concept right into a full comic supply."""
joke_idea: JokeIdea = dspy.InputField()
joke = dspy.OutputField()
class JokeGenerator(dspy.Module):
def __init__(self):
self.query_to_idea = dspy.Predict(QueryToIdea)
self.idea_to_joke = dspy.Predict(IdeaToJoke)
self.query_to_idea.set_lm(lm=dspy.LM("openai/gpt-4.1-mini"))
self.idea_to_joke.set_lm(lm=dspy.LM("openai/gpt-4.1"))
def ahead(self, question):
concept = self.query_to_idea(question=question)
joke = self.idea_to_joke(joke_idea=concept.joke_idea)
return jokeIterative Refinement
You can even implement iterative enchancment the place the LLM displays on and refines its outputs. For instance, we are able to write a refinement module whose context is the output of a earlier LM, and it should act as a suggestions supplier. The primary LM can enter this suggestions and iteratively enhance its response.
iteratively enhance the ultimate joke. (Supply: Creator)
Conditional Branching and Multi-Output Programs
The brokers are related with management flows, and we orchestrate precisely how data flows by means of our system
Generally you need your agent to output a number of variations, after which choose the perfect amongst them. Let’s take a look at an instance of that.
Right here we’ve first outlined a joke decide – it inputs a number of joke concepts, after which picks the index of the perfect joke. This joke is then handed into the following part.
num_samples = 5
class JokeJudge(dspy.Signature):
"""Given an inventory of joke concepts, you have to decide the perfect joke"""
joke_ideas: listing[JokeIdeas] = dspy.InputField()
best_idx: int = dspy.OutputField(
le=num_samples,
ge=1,
description="The index of the funniest joke")
class ConditionalJokeGenerator(dspy.Module):
def __init__(self):
self.query_to_idea = dspy.ChainOfThought(QueryToIdea)
self.decide = dspy.ChainOfThought(JokeJudge)
self.idea_to_joke = dspy.ChainOfThought(IdeaToJoke)
async def ahead(self, question):
# Generate a number of concepts in parallel
concepts = await asyncio.collect(*[
self.query_to_idea.acall(query=query)
for _ in range(num_samples)
])
# Choose and rank concepts
best_idx = (await self.decide.acall(joke_ideas=concepts)).best_idx
# Choose finest concept and generate last joke
best_idea = concepts[best_idx]
# Convert from concept to joke
return await self.idea_to_joke.acall(joke_idea=best_idea)Instrument Calling
LLM functions usually must work together with exterior techniques. That is the place tool-calling steps in. You possibly can think about a software to be any Python perform. You simply want two issues to outline a Python perform as an LLM software:
- An outline of what the perform does
- A listing of inputs and their knowledge varieties
Let’s see an instance of fetching information. We first write a easy Python perform, the place we use Tavily. The perform inputs a search question and fetches current information articles from the final 7 days.
consumer = TavilyClient(api_key=os.getenv("TAVILY_API_KEY"))
def fetch_recent_news(question: str) -> str:
"""Inputs a question string, searches for information and returns prime outcomes."""
response = tavily_client.search(question, search_depth="superior",
matter="information", days=7, max_results=3)
return [x["content"] for x in response["results"]]
Now let’s usedspy.ReAct (or the REasoning and ACTing). The module routinely causes concerning the person’s question, decides when to name which instruments, and incorporates the software outcomes into the ultimate response. Doing that is fairly simple:
class HaikuGenerator(dspy.Signature):
"""
Generates a haiku concerning the newest information on the question.
Additionally create a easy file the place you save the ultimate abstract.
"""
question = dspy.InputField()
abstract = dspy.OutputField(desc="A abstract of the most recent information")
haiku = dspy.OutputField()
program = dspy.ReAct(signature=HaikuGenerator,
instruments=[fetch_recent_news],
max_iters=2)
program.set_lm(lm=dspy.LM("openai/gpt-4.1", temperature=0.7))
pred = program(question="OpenAI")
When the above code runs, the LLM first causes about what the person needs and which software to name (if any). Then it generates the identify of the perform and the arguments to name the perform.
We name the information perform with the generated args, execute the perform to generate the information knowledge. This data is handed again into the LLM. The LLM decides whether or not to name extra instruments, or “end”. If the LLM causes that it has sufficient data to reply the person’s authentic request, it chooses to complete, and generate the reply.
Brokers can take actions whereas producing responses. Every motion the agent can take needs to be effectively outlined so the LLM can work together with it by means of reasoning and appearing.
Superior Instrument Utilization — Scratchpad and File I/O
An evolving commonplace for contemporary functions is to permit LLMs entry to the file system, permitting them to learn and write information, transfer between directories (with acceptable restrictions), grep and search textual content inside information, and even run terminal instructions!
This sample opens a ton of prospects. It transforms the LLM from a passive textual content generator into an lively agent able to performing complicated, multi-step duties immediately inside a person’s setting. For instance, simply displaying the listing of instruments out there to Gemini CLI will reveal a brief however extremely highly effective assortment of instruments.
A fast phrase on MCP Servers
One other new paradigm within the area of agentic techniques are MCP servers. MCPs want their very own devoted article, so I gained’t go over them intimately on this one.
This has rapidly grow to be the industry-standard solution to serve specialised instruments to LLMs. It follows the traditional Consumer-Server structure the place the LLM (a consumer) sends a request to the MCP server, and the MCP server carries out the requested motion, and returns a outcome again to the LLM for downstream processing. MCPs are good for context engineering particular examples since you may declare system immediate codecs, assets, restricted database entry, and so on, to your software.
This repository has an ideal listing of MCP servers that you may examine to make your LLM functions join with all kinds of functions.
Retrieval-Augmented Technology (RAG)
Retrieval Augmented Technology has grow to be a cornerstone of recent AI software improvement. It’s an architectural method that injects exterior, related, and up-to-date data into the Massive Language Fashions (LLMs) that’s contextually related to the person’s question.
RAG pipelines include a preprocessing and an inference-time section. Throughout pre-processing, we course of the reference knowledge corpus and reserve it in a queryable format. Within the inference section, we course of the person question, retrieve related paperwork from our database, and go them into the LLM to generate a response.
The knowledge out there to the brokers can come from a number of sources – exterior database with Retrieval-Augmented Technology (RAG), software calls (like net search), reminiscence techniques, or traditional few-shot examples.
Constructing RAGs is difficult, and there was numerous nice analysis and engineering optimizations which have made life simpler. I made a 17-minute video that covers all of the elements of constructing a dependable RAG pipeline.
Some sensible suggestions for Good RAG
- When preprocessing, generate further metadata per chunk. This may be so simple as “questions this chunk solutions”. When saving the chunks to your database, additionally save the generated metadata!
class ChunkAnnotator(dspy.Signature):
chunk: str = dspy.InputField()
possible_questions: listing[str] = dspy.OutputField(
description="listing of questions that this chunk solutions"
)- Question Rewriting: Straight utilizing the person’s question to do RAG retrieval is usually a nasty concept. Customers write fairly random issues, which can not match the distribution of textual content in your corpus. Question rewriting does what it says – it “rewrites” the question, maybe fixing grammar, spelling errors, contextualizes it with previous dialog, and even provides further key phrases that make querying simpler.
class QueryRewriting(dspy.Signature):
user_query: str = dspy.InputField()
dialog: str = dspy.InputField(
description="The dialog to this point")
modified_query: str = dspy.OutputField(
description="a question contextualizing the person question with the dialog's context and optimized for retrieval search"
)- HYDE or Hypothetical Doc Embedding is a sort of Question Rewriting system. In HYDE, we generate a man-made (or hypothetical) reply from the LLM’s inner data. This response usually incorporates essential key phrases that attempt to immediately match with the solutions database. Vanilla question rewriting is nice for looking a database of questions, and HYDE is nice for looking a database with solutions.
- Hybrid search is sort of all the time higher than purely semantic or purely keyword-based search. For semantic search, I’d use cosine similarity nearest neighbor search with vector embeddings. And for semantic search, use BM25.
- RRF: You possibly can select a number of methods to retrieve paperwork, after which use reciprocal rank fusion to mix them into one unified listing!
- Multi-Hop Search is an possibility to contemplate as effectively when you can afford further latency. Right here, you go the retrieved paperwork again into the LLM to generate new queries, that are used to conduct further searches on the database.
class MultiHopHyDESearch(dspy.Module):
def __init__(self, retriever):
self.generate_queries = dspy.ChainOfThought(QueryGeneration)
self.retriever = retriever
def ahead(self, question, n_hops=3):
outcomes = []
for hop in vary(n_hops): # Discover we loop a number of occasions
# Generate optimized search queries
search_queries = self.generate_queries(
question=question,
previous_jokes=retrieved_jokes
)
# Retrieve utilizing each semantic and key phrase search
semantic_results = self.retriever.semantic_search(
search_queries.semantic_query
)
bm25_results = self.retriever.bm25_search(
search_queries.bm25_query
)
# Fuse outcomes
hop_results = reciprocal_rank_fusion([
semantic_results, bm25_results
])
outcomes.lengthen(hop_results)
return outcomes- Citations: When asking LLM to generate responses from the retrieved paperwork, we are able to additionally ask the LLM to quote references to the paperwork it discovered helpful. This permits the LLM to first generate a plan of the way it’s going to make use of the retrieved content material.
- Reminiscence: In case you are constructing a chatbot, it is very important work out the query of reminiscence. You possibly can think about Reminiscence as a mix of Retrieval and Instrument Calling. A well known system is the Mem0 system. The LLM observes new knowledge and calls instruments to determine if it wants so as to add or modify its current reminiscences. Throughout question-answering, it retrieves related reminiscences utilizing RAG to generate solutions.
Finest Practices and Manufacturing Concerns
This part is just not immediately about Context Engineering, however extra about finest practices to construct LLM apps for manufacturing.
Moreover, techniques should be evaluated with metrics and maintained with observability. Monitoring token utilization, latency, and price to output high quality is a key consideration.
1. Design Analysis First
Earlier than constructing options, determine the way you’ll measure success. This helps scope your software and guides optimization choices.
- In case you can design verifiable or goal rewards, that’s the perfect. (instance: classification duties the place you’ve got a validation dataset)
- If not, are you able to outline features that heuristically consider LLM responses in your use case? (instance: variety of occasions a particular chunk is retrieved given a query)
- If not, are you able to get people to annotate your LLM’s responses?
- If nothing works, use an LLM as a decide to guage responses. Usually, you need to set your analysis job as a comparability examine, the place the Choose receives a number of responses produced utilizing totally different hyperparameters/prompts, and the decide should rank which of them are the perfect.
3. Use Structured Outputs Nearly In every single place
All the time favor structured outputs over free-form textual content. It makes your system extra dependable and simpler to debug. You possibly can add validation and retries as effectively!
4. Design for failure
When designing prompts or dspy modules, be sure to all the time contemplate “what occurs if issues go mistaken?”
Like every good software program, slicing down error states and failing with swagger is the perfect situation.
5. Monitor Every little thing
DSpy integrates with MLflow to trace:
- Particular person prompts handed into the LLM and their responses
- Token utilization and prices
- Latency per module
- Success/failure charges
- Mannequin efficiency over time
Langfuse, Logfire are equally nice options.
Outro
Context engineering represents a paradigm shift from easy immediate engineering to constructing complete and modular LLM functions.
The DSPy framework supplies the instruments and abstractions wanted to implement these patterns systematically. As LLM capabilities proceed to evolve, context engineering will grow to be more and more essential for constructing functions that successfully leverage the facility of huge language fashions.
To observe the total video course on which this text is predicated, please go to this YouTube hyperlink.
To entry the total GitHub repo, go to:
https://github.com/avbiswas/context-engineering-dspy
References
Creator’s YouTube channel: https://www.youtube.com/@avb_fj
Creator’s Patreon: www.patreon.com/NeuralBreakdownwithAVB
Creator’s Twitter (X) account: https://x.com/neural_avb
Full Context Engineering video course: https://youtu.be/5Bym0ffALaU
Github Hyperlink: https://github.com/avbiswas/context-engineering-dspy
