MODULE 5 DEEP DIVE · CONTAINERS FOR GENAI & AGENTIC AI
The knobs nobody explained
The lab's defaults were never explained — this deck opens them
Gourav Shah · School of DevOps & AI · Deep Dive (Part 2)
M5-DD·01
Cut the index card too small and the fact loses its command
A chunk boundary mid-idea splits a fact from the step that follows it
M5-DD·02
Cut it too large and four procedures blur into one average
An oversized chunk compresses unrelated procedures into one blurry vector
M5-DD·03
Overlap photocopies the tail onto the next card
The tail of chunk 1 is copied onto chunk 2 — no fact is ever orphaned
M5-DD·04
This corpus split into exactly 2 chunks, and that's why
chunk_size=500 on this ~800-char corpus produces exactly 2 chunks
M5-DD·05
Top-k decides how many cards land on the model's desk
k=3 always retrieves the three nearest chunks, relevant or not
M5-DD·06
k=3 requested, 2 returned — you're already retrieving everything
Only 2 chunks exist here, so k=3 can't filter anything at all
M5-DD·07
Distance, not similarity — and lower means closer
ChromaDB ranks by distance, not similarity — smaller number wins
M5-DD·08
Same ranking here, but only because the vectors sit on the unit circle
On the unit circle, L2 distance and cosine angle agree exactly
M5-DD·09
The embedder decides what "similar" means — the LLM never gets a vote
The LLM never sees the corpus until retrieval hands it a slice
M5-DD·10
Two embedding models never share a coordinate system
Two embedding models never share a coordinate system
M5-DD·11
Four pieces have to fit inside one 4096-token ceiling
Four pieces compete for one fixed 4096-token ceiling
M5-DD·12
At this lab's scale, the prompt uses under 15% of the ceiling
This prompt uses under 15% of the 4096-token ceiling
M5-DD·13
Ollama keeps the tail and drops the front, verified live
Verified live: Ollama keeps the tail and drops the front
M5-DD·14
Two answers wrong for two different reasons
Two wrong answers, two different root causes to check first
M5-DD·15
Naive RAG never asks "is the 2nd-best chunk actually better?"
Naive RAG never re-ranks — top distance wins, straight to the model
M5-DD·16
ChromaDB's own API confirms the metric, not a guess
ChromaDB's own API confirms the metric — no override was ever set
M5-DD·17
0.6956 and 1.0968 — the numbers behind "Found 2 relevant chunks"
The real distances behind "Found 2 relevant chunks"
M5-DD·18
Three sentences, three vectors, norm 1.000000 every time
Three vectors, norm 1.000000 every time — measured, not assumed
M5-DD·19
Same corpus, three chunk sizes, three different stories
Same corpus, three chunk sizes, three very different outcomes
M5-DD·20
MODULE 5 DEEP DIVE · TAKEAWAYS
On a corpus this small, even the worst chunking still works
Next: apply these levers to a real corpus. · Gourav Shah · School of DevOps & AI
M5-DD·21