ragit 0.4.5

git-like rag pipeline
Documentation 
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
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168

import os
import re
import shutil
from utils import (
    cargo_run,
    goto_root,
    ls_recursive,
    mk_and_cd_tmp_dir,
    rand_word,
    write_string,
)

def ii():
    goto_root()
    mk_and_cd_tmp_dir()

    # step 1: create a local knowledge-base
    os.mkdir("base")
    os.chdir("base")
    shutil.copytree("../../src", "src")
    os.chdir("src")

    if ".ragit" in os.listdir():
        shutil.rmtree(".ragit")

    cargo_run(["init"])
    cargo_run(["config", "--set", "model", "dummy"])
    cargo_run(["config", "--set", "chunk_size", "512"])
    cargo_run(["config", "--set", "slide_len", "128"])
    cargo_run(["config", "--set", "enable_ii", "false"])
    cargo_run(["add", *(ls_recursive("rs") + ls_recursive("txt"))])

    cargo_run(["build"])
    # `rag build` builds an inverted-index
    assert cargo_run(["ii-status"], stdout=True).strip() == "complete"
    # we have to reset ii in order to run `ii_worker`
    cargo_run(["ii-reset"])
    assert cargo_run(["ii-status"], stdout=True).strip() == "not initialized"
    ii_worker()
    os.chdir("..")

    # step 2: clone remote knowledge-bases
    for (url, dir) in [
        ("https://ragit.baehyunsol.com/sample/git", "git"),
        ("https://ragit.baehyunsol.com/sample/ragit", "ragit"),
        ("https://ragit.baehyunsol.com/sample/rustc", "rustc"),
    ]:
        cargo_run(["clone", url, dir])
        os.chdir(dir)
        cargo_run(["config", "--set", "enable_ii", "false"])
        cargo_run(["config", "--set", "model", "dummy"])
        ii_worker()
        os.chdir("..")

def ii_worker():
    # Strategy:
    # 1. There are an arbitrary number of chunks that contain `terms`.
    #    But let's make sure that some terms appear very often and some
    #    terms appear very rarely, for better test coverage.
    # 2. It runs `tfidf <term>` for each `term` in `terms`.
    #    Each time, it records the rank of the chunks.
    # 3. After running `ii-build`, it runs `tfidf <term>` again.
    #    Again it records the rank of the chunks.
    # 4. Result from 2 is the answer and result from 3 is an 'approximation'.
    #    It's goal is to make sure that the approximation is close enough.
    # 5. Rules
    #    - If answer has >= 10 chunks and approximation has >= 10 chunks,
    #      - The top 3 of each set must be included in the top 10 of the other.
    #    - If answer has < 10 chunks and approximation has < 10 chunks,
    #      - The set of the chunks must be the same.
    #    - Otherwise,
    #      - Error
    terms = generate_terms()
    answers = {}
    approximations = {}

    for term in terms:
        uids = cargo_run(["tfidf", term, "--uid-only", "--limit=10"], stdout=True)
        uids = [uid for uid in uids.split("\n") if re.match(r"^[0-9a-z]{64}$", uid)]
        assert len(uids) <= 10
        answers[term] = uids

    cargo_run(["ii-build"])
    cargo_run(["check"])
    assert cargo_run(["ii-status"], stdout=True).strip() == "complete"
    cargo_run(["config", "--set", "enable_ii", "true"])

    for term in terms:
        uids = cargo_run(["tfidf", term, "--uid-only", "--limit=10"], stdout=True)
        uids = [uid for uid in uids.split("\n") if re.match(r"^[0-9a-z]{64}$", uid)]
        assert len(uids) <= 10
        approximations[term] = uids

    for term in terms:
        answer = answers[term]
        approximation = approximations[term]

        try:
            if len(answer) == 10:
                if len(approximation) == 10:
                    for i in range(3):
                        if answer[i] not in approximation:
                            raise AssertionError(f"answer[{i}] not in approximation")

                        if approximation[i] not in answer:
                            raise AssertionError(f"approximation[{i}] not in answer")

                else:
                    raise AssertionError(f"len(answer) == 10, len(approximation) == {len(approximation)}")

            else:
                if len(approximation) == 10:
                    raise AssertionError(f"len(answer) == {len(answer)}, len(approximation) == 10")

                elif set(answer) != set(approximation):
                    raise AssertionError(f"set(answer) != set(approximation)")

        except AssertionError as e:
            raise AssertionError(f"tfidf result on term '{term}' is not close enough. error: `{e}`, answer: {answer}, approximation: {approximation}")

    # incremental update of ii
    write_string("self-introduction.txt", "Hi, my name is baehyunsol.")
    cargo_run(["add", "self-introduction.txt"])
    assert cargo_run(["ii-status"], stdout=True).strip() == "complete"
    cargo_run(["build"])
    cargo_run(["check"])
    assert cargo_run(["ii-status"], stdout=True).strip() == "complete"
    assert "self-introduction.txt" in cargo_run(["tfidf", "Hi, my name is baehyunsol."], stdout=True)

def generate_terms():
    dictionary = []
    words = set()

    for line in cargo_run(["ls-terms"], stdout=True).split("\n"):
        if (r := re.match(r"^\s*\"([0-9a-zA-Z가-힣]+)\"\:\s*(\d+)$", line)) is not None:
            dictionary.append((r.group(1), int(r.group(2))))
            words.add(r.group(1))

    dictionary.sort(key=lambda x: x[0])  # For deterministic result
    dictionary.sort(key=lambda x: x[1], reverse=True)
    dictionary = [term for term, _ in dictionary]

    if len(dictionary) < 500:
        raise Exception("The dataset is not big enough. Please make sure that there are more than 500 terms.")

    very_frequent_terms = dictionary[:5]
    frequent_terms = dictionary[len(dictionary) // 20:len(dictionary) // 20 + 5]
    less_frequent_terms = dictionary[len(dictionary) // 2:len(dictionary) // 2 + 5]
    rare_terms = dictionary[-5:]
    never_terms = []

    while len(never_terms) < 5:
        term = rand_word()

        if term not in words:
            never_terms.append(term)

    return [
        *frequent_terms,
        *less_frequent_terms,
        *rare_terms,
        *never_terms,
        " ".join(frequent_terms),
        " ".join(less_frequent_terms),
        " ".join(rare_terms),
        *[" ".join([very_frequent_terms[i], less_frequent_terms[i], rare_terms[i]]) for i in range(5)],
        *[" ".join([very_frequent_terms[i], less_frequent_terms[i]]) for i in range(5)],
    ]