implementation

LLM-Driven Forecasting and DSPy Optimization

Inspect the original prompt language first, then copy or adapt it once you know how it fits your workflow.

Linked challenge: Adaptive LLM Time Series Forecasting with DSPy & W&B

Format

Text-first

Lines

Sections

Linked challenge

Adaptive LLM Time Series Forecasting with DSPy & W&B

Prompt source

Original prompt text with formatting preserved for inspection.

1 lines

1 sections

No variables

0 checklist items

Implement the core forecasting logic. Using the semantic abstractions generated in the previous step, prompt the LLM (GPT-5 Pro via Together AI) to generate future predictions for a given forecast horizon. Apply DSPy to systematically optimize the LLM's forecasting prompts and reasoning steps. This might involve few-shot examples, chain-of-thought prompting, or custom DSPy modules. Track the performance of different DSPy-optimized prompts using Weights & Biases.

Adaptation plan

Keep the source stable, then change the prompt in a predictable order so the next run is easier to evaluate.

Keep stable

Hold the task contract and output shape stable so generated implementations remain comparable.

Tune next

Update libraries, interfaces, and environment assumptions to match the stack you actually run.

Verify after

Test failure handling, edge cases, and any code paths that depend on hidden context or secrets.

Prompt diagnostics

Variables

Lists

Code blocks

Purpose

implementation

This prompt is mostly narrative and instruction-driven, so adapt examples and output constraints before you rewrite the structure.

Linked challenge

Adaptive LLM Time Series Forecasting with DSPy & W&B

This challenge focuses on building an advanced time series forecasting system that leverages the power of Large Language Models (LLMs) like GPT-5 Pro. Instead of traditional statistical or deep learning models, you will design a system where an LLM performs forecasting by first generating "semantic abstractions" of the time series data (e.g., identifying underlying trends, seasonality, significant events, and causal factors from textual descriptions or metadata). These abstractions will then guide the LLM's forecasting process, allowing for more interpretable and adaptable predictions. The system should also incorporate principles of continuous learning, demonstrating how the forecasting model can adapt to new data patterns or updated domain knowledge without requiring full retraining from scratch. You will use DSPy to programmatically optimize the LLM's prompting and reasoning steps for forecasting and integrate Weights & Biases for experiment tracking, model evaluation, and monitoring the adaptive learning process. Apache Airflow will orchestrate the data ingestion, semantic abstraction generation, forecasting, and adaptation pipeline.

Open challenge

Related prompts

Browse library

Architecting the Adaptive Forecasting Pipeline

planning

Implementing Semantic Abstraction with GPT-5 Pro

implementation

Implementing Continuous Learning and Adaptation

implementation