You cannot embed a whole manual as one item and search it well. It is split into pieces, and the size and shape of those pieces decide whether a question can find its answer. Chunking is the one decision that most retrieval quality is made of.
About 1 min of reading
Chunk a small document by heading.
The document
Before you run it, say out loud what it will print. Then open the answer.
The chunks
Before you run it, say out loud what it will print. Then open the answer.
Before you write anything, put these lines back in the right order. Click a line to move it up one step. This support step is easier than writing code from scratch, and it is here to make the last example stick.
Read the first example one line at a time. A line that prints shows what appears on screen; the rest set things up quietly.
Prepend the section heading to every chunk in that section. Now a chunk says which section it belongs to, and a question naming that section matches strongly. That one change fixes most weak retrieval.
Every chunk needs a stable identifier and a pointer back to the document. Without it you cannot cite the source, and a citation is what makes the answer checkable.
A chunk that holds one idea answers one question cleanly. A chunk that holds six topics dilutes the match, because the embedding now describes all six.
A chunk that is one sentence loses the sentence before it that named the subject. Overlap, or a heading carried into each chunk, is what keeps the reference.
Split on paragraph or heading boundaries so a chunk is a whole thought. Splitting at a fixed character count cuts a list in half and creates a chunk that matches nothing well.
Each one is defined here on this site. Open the entry for the full definition and the other words it points at.