Docs/Core: Low-Resource Setups, SkyPilot Option, and No-Recompute (#45)

* docs: add SkyPilot template and instructions for running embeddings/index build on cloud GPU

* docs: add low-resource note in README; point to config guide; suggest OpenAI embeddings, SkyPilot remote build, and --no-recompute

* docs: consolidate low-resource guidance into config guide; README points to it

* cli: add --no-recompute and --no-recompute-embeddings flags; docs: clarify HNSW requires --no-compact when disabling recompute

* docs: dedupe recomputation guidance; keep single Low-resource setups section

* sky: expand leann-build.yaml with configurable params and flags (backend, recompute, compact, embedding options)

* hnsw: auto-disable compact when --no-recompute is used; docs: expand SkyPilot with -e overrides and copy-back example

* docs+sky: simplify SkyPilot flow (auto-build on launch, rsync copy-back); clarify HNSW auto non-compact when no-recompute

* feat: auto compact for hnsw when recompute

* reader: non-destructive portability (relative hints + fallback); fix comments; sky: refine yaml

* cli: unify flags to --recompute/--no-recompute for build/search/ask; docs: update references

* chore: remove

* hnsw: move pruned/no-recompute assertion into backend; api: drop global assertion; docs: will adjust after benchmarking

* cli: use argparse.BooleanOptionalAction for paired flags (--recompute/--compact) across build/search/ask

* docs: a real example on recompute

* benchmarks: fix and extend HNSW+DiskANN recompute vs no-recompute; docs: add fresh numbers and DiskANN notes

* benchmarks: unify HNSW & DiskANN into one clean script; isolate groups, fixed ports, warm-up, param complexity

* docs: diskann recompute

* core: auto-cleanup for LeannSearcher/LeannChat (__enter__/__exit__/__del__); ensure server terminate/kill robustness; benchmarks: use searcher.cleanup(); docs: suggest uv run

* fix: hang on warnings

* docs: boolean flags

* docs: leann help
This commit is contained in:
Andy Lee
2025-08-15 12:03:19 -07:00
committed by GitHub
parent 00eeadb9dd
commit db3c63c441
12 changed files with 529 additions and 101 deletions

View File

@@ -97,29 +97,23 @@ ollama pull nomic-embed-text
```
### DiskANN
**Best for**: Performance-critical applications and large datasets - **Production-ready with automatic graph partitioning**
**Best for**: Large datasets, especially when you want `recompute=True`.
**How it works:**
- **Product Quantization (PQ) + Real-time Reranking**: Uses compressed PQ codes for fast graph traversal, then recomputes exact embeddings for final candidates
- **Automatic Graph Partitioning**: When `is_recompute=True`, automatically partitions large indices and safely removes redundant files to save storage
- **Superior Speed-Accuracy Trade-off**: Faster search than HNSW while maintaining high accuracy
**Key advantages:**
- **Faster search** on large datasets (3x+ speedup vs HNSW in many cases)
- **Smart storage**: `recompute=True` enables automatic graph partitioning for smaller indexes
- **Better scaling**: Designed for 100k+ documents
**Trade-offs compared to HNSW:**
- **Faster search latency** (typically 2-8x speedup)
- **Better scaling** for large datasets
-**Smart storage management** with automatic partitioning
-**Better graph locality** with `--ldg-times` parameter for SSD optimization
- ⚠️ **Slightly larger index size** due to PQ tables and graph metadata
**Recompute behavior:**
- `recompute=True` (recommended): Pure PQ traversal + final reranking - faster and enables partitioning
- `recompute=False`: PQ + partial real distances during traversal - slower but higher accuracy
```bash
# Recommended for most use cases
--backend-name diskann --graph-degree 32 --build-complexity 64
# For large-scale deployments
--backend-name diskann --graph-degree 64 --build-complexity 128
```
**Performance Benchmark**: Run `python benchmarks/diskann_vs_hnsw_speed_comparison.py` to compare DiskANN and HNSW on your system.
**Performance Benchmark**: Run `uv run benchmarks/diskann_vs_hnsw_speed_comparison.py` to compare DiskANN and HNSW on your system.
## LLM Selection: Engine and Model Comparison
@@ -273,24 +267,114 @@ Every configuration choice involves trade-offs:
The key is finding the right balance for your specific use case. Start small and simple, measure performance, then scale up only where needed.
## Deep Dive: Critical Configuration Decisions
## Low-resource setups
### When to Disable Recomputation
If you dont have a local GPU or builds/searches are too slow, use one or more of the options below.
LEANN's recomputation feature provides exact distance calculations but can be disabled for extreme QPS requirements:
### 1) Use OpenAI embeddings (no local compute)
Fastest path with zero local GPU requirements. Set your API key and use OpenAI embeddings during build and search:
```bash
--no-recompute # Disable selective recomputation
export OPENAI_API_KEY=sk-...
# Build with OpenAI embeddings
leann build my-index \
--embedding-mode openai \
--embedding-model text-embedding-3-small
# Search with OpenAI embeddings (recompute at query time)
leann search my-index "your query" \
--recompute
```
**Trade-offs**:
- **With recomputation** (default): Exact distances, best quality, higher latency, minimal storage (only stores metadata, recomputes embeddings on-demand)
- **Without recomputation**: Must store full embeddings, significantly higher memory and storage usage (10-100x more), but faster search
### 2) Run remote builds with SkyPilot (cloud GPU)
Offload embedding generation and index building to a GPU VM using [SkyPilot](https://skypilot.readthedocs.io/en/latest/). A template is provided at `sky/leann-build.yaml`.
```bash
# One-time: install and configure SkyPilot
pip install skypilot
# Launch with defaults (L4:1) and mount ./data to ~/leann-data; the build runs automatically
sky launch -c leann-gpu sky/leann-build.yaml
# Override parameters via -e key=value (optional)
sky launch -c leann-gpu sky/leann-build.yaml \
-e index_name=my-index \
-e backend=hnsw \
-e embedding_mode=sentence-transformers \
-e embedding_model=Qwen/Qwen3-Embedding-0.6B
# Copy the built index back to your local .leann (use rsync)
rsync -Pavz leann-gpu:~/.leann/indexes/my-index ./.leann/indexes/
```
### 3) Disable recomputation to trade storage for speed
If you need lower latency and have more storage/memory, disable recomputation. This stores full embeddings and avoids recomputing at search time.
```bash
# Build without recomputation (HNSW requires non-compact in this mode)
leann build my-index --no-recompute --no-compact
# Search without recomputation
leann search my-index "your query" --no-recompute
```
When to use:
- Extreme low latency requirements (high QPS, interactive assistants)
- Read-heavy workloads where storage is cheaper than latency
- No always-available GPU
Constraints:
- HNSW: when `--no-recompute` is set, LEANN automatically disables compact mode during build
- DiskANN: supported; `--no-recompute` skips selective recompute during search
Storage impact:
- Storing N embeddings of dimension D with float32 requires approximately N × D × 4 bytes
- Example: 1,000,000 chunks × 768 dims × 4 bytes ≈ 2.86 GB (plus graph/metadata)
Converting an existing index (rebuild required):
```bash
# Rebuild in-place (ensure you still have original docs or can regenerate chunks)
leann build my-index --force --no-recompute --no-compact
```
Python API usage:
```python
from leann import LeannSearcher
searcher = LeannSearcher("/path/to/my-index.leann")
results = searcher.search("your query", top_k=10, recompute_embeddings=False)
```
Trade-offs:
- Lower latency and fewer network hops at query time
- Significantly higher storage (10100× vs selective recomputation)
- Slightly larger memory footprint during build and search
Quick benchmark results (`benchmarks/benchmark_no_recompute.py` with 5k texts, complexity=32):
- HNSW
```text
recompute=True: search_time=0.818s, size=1.1MB
recompute=False: search_time=0.012s, size=16.6MB
```
- DiskANN
```text
recompute=True: search_time=0.041s, size=5.9MB
recompute=False: search_time=0.013s, size=24.6MB
```
Conclusion:
- **HNSW**: `no-recompute` is significantly faster (no embedding recomputation) but requires much more storage (stores all embeddings)
- **DiskANN**: `no-recompute` uses PQ + partial real distances during traversal (slower but higher accuracy), while `recompute=True` uses pure PQ traversal + final reranking (faster traversal, enables build-time partitioning for smaller storage)
**Disable when**:
- You have abundant storage and memory
- Need extremely low latency (< 100ms)
- Running a read-heavy workload where storage cost is acceptable
## Further Reading