Skip to content
anythingai.space
Frameworks & Toolsintermediate13 min · 8 steps

LlamaIndex and DSPy

Two specialists: LlamaIndex for data-shaped work (parsing, indexing, event-driven workflows) and DSPy for turning prompts into compiled, optimisable programs.

Full chapter: LlamaIndex
DATA-SHAPEDPROMPT-SHAPED1,000 PDFstables · scansRule-based parserIndexHand-tuned prompt100 linesNew modelshipped this monthBroken output
dataerrorrequesttool callresponseheavy work
1 / 8
ProblemStep 1 of 8

Two different hard parts

Some LLM apps are data-shaped: the hard part is getting 1,000 messy PDFs (tables, scans, odd layouts) parsed, chunked and indexed. A rule-based parser mangles them, and no prompt can fix bad input.

Others are prompt-shaped: a 100-line hand-tuned prompt works on today's model and breaks when a new one ships. In September 2026 alone, Anthropic shipped Fable 5.1, Opus 5.5 and Sonnet 5.5, and OpenAI shipped several GPT-6 models.

Up nextLlamaIndex targets the first problem, DSPy the second.

Lesson 2 of 6 in Frameworks & Tools