offlineaid-pack-builder/Makefile at main · helenkwok/offlineaid-pack-builder · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
INPUTS ?= ./examples/accc-scams
BUILD_DIR ?= ./build/extracted
OUT_DIR ?= ./out
PACK_NAME ?= scam-resilience-au
COMPILER_MODEL ?= offlineaid-compiler
OCR_MODEL ?= gemma4-26b-ocr
OLLAMA_BASE_URL ?= http://localhost:11434/v1
DATASET_DIR ?= ./build/dataset
TRAIN_EN ?= ./build/train_en.jsonl
CORPUS_DIR ?= ./build/extracted
LANGS ?= zh ar
EVAL_DIR ?= ./eval
EVAL_DB  ?= $(OUT_DIR)/$(PACK_NAME).db

.PHONY: all pack clean verify test scrape translate validate upload eval eval-retrieval eval-mac eval-device eval-write

all: pack

pack:
	@mkdir -p $(BUILD_DIR) $(OUT_DIR)
	@T0=$$(date +%s); \
	uv run --extra ocr python extract.py "$(INPUTS)" --output-dir "$(BUILD_DIR)" --model "$(OCR_MODEL)" --base-url "$(OLLAMA_BASE_URL)" && \
	uv run --extra agent python agent.py --input "$(BUILD_DIR)" --output "$(OUT_DIR)/$(PACK_NAME).oapack.zip" --model "$(COMPILER_MODEL)" --base-url "$(OLLAMA_BASE_URL)" && \
	uv run python packager.py verify "$(OUT_DIR)/$(PACK_NAME).oapack.zip" && \
	T1=$$(date +%s); \
	echo "{\"wall_time_s\": $$(($$T1-$$T0))}" > "$(OUT_DIR)/$(PACK_NAME).build-time.json"

clean:
	rm -rf $(BUILD_DIR) $(OUT_DIR)

verify:
	uv run python packager.py verify "$(OUT_DIR)/$(PACK_NAME).oapack.zip"

test:
	uv run --extra agent --extra ocr --extra scrape pytest -q

scrape:
	@mkdir -p $(BUILD_DIR)
	uv run --extra scrape python scrape.py --output-dir "$(BUILD_DIR)"

validate:
	uv run --extra agent python validate.py \
	  --input "$(TRAIN_EN)" \
	  --corpus-dir "$(CORPUS_DIR)" \
	  --threshold 0.6

translate:
	uv run --extra agent python translate.py \
	  --input "$(TRAIN_EN)" \
	  --output-dir ./build \
	  --langs $(LANGS)

upload:
	@mkdir -p $(DATASET_DIR)
	cp ./build/train_en.jsonl ./build/train_zh.jsonl ./build/train_ar.jsonl \
	   ./build/train_multilingual.jsonl $(DATASET_DIR)/
	uv run --extra upload kaggle datasets version \
	  -p $(DATASET_DIR) \
	  -m "v2: ~2500 records en+zh+ar"

# ─── Eval harness ─────────────────────────────────────────────────────────
eval: eval-retrieval eval-mac eval-device eval-write

eval-retrieval:
	uv run pytest $(EVAL_DIR)/test_retrieval.py -v --junitxml=$(EVAL_DIR)/.retrieval.xml

eval-mac:
	uv run pytest $(EVAL_DIR)/test_mac_latency.py -v --junitxml=$(EVAL_DIR)/.mac.xml

eval-device:
	uv run pytest $(EVAL_DIR)/test_device_e2e.py -v --junitxml=$(EVAL_DIR)/.device.xml

eval-write:
	uv run python $(EVAL_DIR)/perf_writer.py \
	  --retrieval $(EVAL_DIR)/.retrieval.xml \
	  --mac $(EVAL_DIR)/.mac.xml \
	  --device $(EVAL_DIR)/.device.xml \
	  --output $(EVAL_DIR)/PERF.md