Skip to content

Artifact Archive

MinioArchive implements S3-compatible artifact and corpus support.

This API is important for:

  • moving local corpus payloads into remotely readable object storage
  • task-scoped output references
  • signed URL generation
  • fallback behavior between local file paths and remote object locators

lingo.bucket.archive

S3-compatible archive for storage and retrieval.

MinioArchive

Bases: Archive

S3-compatible implementation of the Archive interface.

Source code in lingo/bucket/archive.py
 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
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
class MinioArchive(Archive):
    """S3-compatible implementation of the Archive interface."""

    def __init__(
        self,
        bucket: str = "default",
        prefix: str = "",
        endpoint: Optional[str] = None,
        access_key: Optional[str] = None,
        secret_key: Optional[str] = None,
        secure: bool = False,
        root_dir: Optional[str] = None,
    ):
        self.bucket = bucket
        self.prefix = prefix
        self.endpoint = (endpoint or os.getenv("MINIO_ENDPOINT", "http://localhost:9000")).rstrip("/")
        self.access_key = access_key or os.getenv("MINIO_ACCESS_KEY") or os.getenv("MINIO_ROOT_USER") or "minioadmin"
        self.secret_key = secret_key or os.getenv("MINIO_SECRET_KEY") or os.getenv("MINIO_ROOT_PASSWORD") or "minioadmin"
        self.secure = secure
        self._endpoint_url = self._normalize_endpoint_url(self.endpoint, secure=self.secure)
        base = Path(root_dir) if root_dir else Path(tempfile.gettempdir()) / "lingo_archive"
        self.root_dir = base / bucket
        self.root_dir.mkdir(parents=True, exist_ok=True)
        self._s3_client = None
        # Backward-compatible alias for tests or integrations still patching this attribute.
        self._minio_client = None

        try:
            self._s3_client = boto3.client(
                "s3",
                endpoint_url=self._endpoint_url,
                aws_access_key_id=self.access_key,
                aws_secret_access_key=self.secret_key,
            )
            self._minio_client = self._s3_client
            try:
                self._s3_client.head_bucket(Bucket=self.bucket)
            except ClientError as exc:
                error_code = exc.response.get("Error", {}).get("Code", "")
                if error_code in {"404", "NoSuchBucket", "NotFound"}:
                    self._s3_client.create_bucket(Bucket=self.bucket)
                else:
                    raise
        except Exception:
            print(f'S3 client failed to start with endpoint "{self._endpoint_url}".')
            raise

    @staticmethod
    def _normalize_endpoint_url(endpoint: str, *, secure: bool) -> str:
        parsed = urlparse(endpoint)
        if parsed.scheme:
            return endpoint.rstrip("/")
        scheme = "https" if secure else "http"
        return f"{scheme}://{endpoint.strip('/')}"

    @property
    def normalized_prefix(self) -> str:
        value = (self.prefix or "").strip("/")
        return f"{value}/" if value else ""

    @property
    def _endpoint_host(self) -> str:
        parsed = urlparse(self._endpoint_url)
        return parsed.netloc.split("@")[-1]

    def has_upload_client(self) -> bool:
        return self._s3_client is not None

    @staticmethod
    def _extension_for_content_type(content_type: Optional[str]) -> str:
        if not content_type:
            return ""
        ext = mimetypes.guess_extension(content_type, strict=False)
        return ext or ""

    def _build_object_key(
        self,
        *,
        path: Optional[str] = None,
        task_id: Optional[str] = None,
        content_type: Optional[str] = None,
    ) -> str:
        if path:
            clean = path.lstrip("/")
            if task_id:
                return f"{self.normalized_prefix}{task_id}/{clean}" if self.normalized_prefix else f"{task_id}/{clean}"
            return f"{self.normalized_prefix}{clean}" if self.normalized_prefix else clean

        generated_name = f"{uuid.uuid4().hex}{self._extension_for_content_type(content_type)}"
        if task_id:
            return f"{self.normalized_prefix}{task_id}/{generated_name}" if self.normalized_prefix else f"{task_id}/{generated_name}"
        return f"{self.normalized_prefix}{generated_name}" if self.normalized_prefix else generated_name

    def _local_path_from_object_key(self, object_key: str) -> Path:
        return self.root_dir / object_key.lstrip("/")

    def _resolve_path(self, relative_path: str) -> Path:
        key = self._build_object_key(path=relative_path)
        safe_key = key.lstrip("/")
        return self.root_dir / safe_key

    def _parse_minio_locator(self, locator: str) -> Optional[tuple[str, str]]:
        parsed = urlparse(locator)
        if parsed.scheme in {"s3", "minio"} and parsed.netloc and parsed.path:
            return parsed.netloc, parsed.path.lstrip("/")

        if parsed.scheme in {"http", "https"} and parsed.netloc:
            endpoint_host = self._endpoint_host
            locator_host = parsed.netloc.split("@")[-1]
            if locator_host != endpoint_host:
                return None
            path = parsed.path.lstrip("/")
            if not path or "/" not in path:
                return None
            bucket, object_key = path.split("/", 1)
            return bucket, object_key

        return None

    def _public_url_for_object(self, bucket: str, object_key: str) -> str:
        if self._s3_client is not None:
            return self._s3_client.generate_presigned_url(
                "get_object",
                Params={"Bucket": bucket, "Key": object_key},
                ExpiresIn=24 * 60 * 60,
            )
        return f"{self._endpoint_url}/{bucket}/{object_key}"

    def _object_key_from_reference(self, reference: Reference[Any]) -> str:
        if reference.object_key:
            return reference.object_key.lstrip("/")
        ref_path = Path(reference.path)
        try:
            return ref_path.relative_to(self.root_dir).as_posix()
        except ValueError:
            return ref_path.name

    async def create_task_reference(
        self,
        *,
        task_id: str,
        content_type: Optional[str] = None,
        path_hint: Optional[str] = None,
    ) -> Reference[bytes]:
        object_key = self._build_object_key(path=path_hint, task_id=task_id, content_type=content_type)
        local_path = self._local_path_from_object_key(object_key)
        return Reference.from_path(local_path, content_type=content_type, object_key=object_key)

    async def prepare_corpus_for_dispatch(self, corpus: Corpus[Any], *, task_id: str) -> Corpus[Any]:
        """Ensure local corpus payload is remotely readable for distributed dispatch."""
        source_path = corpus.source_path
        if not source_path:
            return corpus

        local_path = Path(source_path)
        if not local_path.exists():
            return corpus

        reference = await self.create_task_reference(
            task_id=task_id,
            content_type=corpus.content_type,
            path_hint=local_path.name,
        )
        reference.dump_file(local_path, content_type=corpus.content_type)
        await self.ensure_reference_uploaded(reference)
        public_url = await self.get_public_url(reference)
        return Corpus[Any](
            content_type=corpus.content_type,
            source_url=public_url,
            object_data=corpus.object_data,
            bytes_data=None,
            source_path=None,
        )

    async def corpus_from_minio(
        self,
        *,
        object_key: str,
        bucket: Optional[str] = None,
        content_type: Optional[str] = None,
        copy_to_task: bool = False,
        task_id: Optional[str] = None,
    ) -> Corpus[bytes]:
        """Create a corpus pointing to an existing S3-compatible object.

        By default this does not copy data into `{prefix}/{task_id}`. Set
        `copy_to_task=True` to materialize a task-scoped duplicate.
        """
        source_bucket = bucket or self.bucket
        source_key = object_key.lstrip("/")

        if not copy_to_task:
            return Corpus[bytes](
                content_type=content_type,
                source_url=self._public_url_for_object(source_bucket, source_key),
            )

        if not task_id:
            raise ValueError("task_id is required when copy_to_task=True")
        if self._s3_client is None:
            raise RuntimeError("copy_to_task requires an initialized S3 client")

        target_ref = await self.create_task_reference(
            task_id=task_id,
            content_type=content_type,
            path_hint=Path(source_key).name,
        )
        target_key = target_ref.object_key or self._object_key_from_reference(target_ref)

        try:
            self._s3_client.copy_object(
                Bucket=self.bucket,
                Key=target_key,
                CopySource={"Bucket": source_bucket, "Key": source_key},
            )
        except Exception:
            source_url = self._public_url_for_object(source_bucket, source_key)
            with urlopen(source_url, timeout=20) as response:
                data = response.read()
            target_ref.dump_content(data, content_type=content_type)
            await self.ensure_reference_uploaded(target_ref)

        return Corpus[bytes](
            content_type=content_type,
            source_url=await self.get_public_url(target_ref),
        )

    async def ensure_reference_uploaded(self, reference: Reference[Any]) -> None:
        """Upload local reference file to S3-compatible storage when configured."""
        if self._s3_client is None:
            return

        local_path = Path(reference.path)
        if not local_path.exists():
            return

        object_key = self._object_key_from_reference(reference)
        extra_args = {"ContentType": reference.content_type or "application/octet-stream"}
        self._s3_client.upload_file(
            str(local_path),
            self.bucket,
            object_key,
            ExtraArgs=extra_args,
        )

    async def corpus_from_url(
        self,
        url: str,
        content_type: Optional[str] = None,
    ) -> Corpus[bytes]:
        """Load a corpus from a URL."""
        local_candidate = Path(url)
        if local_candidate.exists():
            return Corpus.from_file(local_candidate, content_type=content_type)

        parsed = urlparse(url)
        if parsed.scheme in ("", "file"):
            local_path = Path(parsed.path if parsed.scheme == "file" else url)
            return Corpus.from_file(local_path, content_type=content_type)

        locator = self._parse_minio_locator(url)
        if locator is not None:
            bucket, object_key = locator
            return await self.corpus_from_minio(
                bucket=bucket,
                object_key=object_key,
                content_type=content_type,
                copy_to_task=False,
            )

        with urlopen(url, timeout=10) as response:
            data = response.read()
        return Corpus[bytes](content_type=content_type, bytes_data=data)

    async def get_reference(
        self,
        path: Optional[str] = None,
        content_type: Optional[str] = None,
        *,
        task_id: Optional[str] = None,
    ) -> Reference[bytes]:
        """Get a reference to a storage location."""
        object_key = self._build_object_key(path=path, task_id=task_id, content_type=content_type)
        full_path = self._local_path_from_object_key(object_key)
        return Reference.from_path(full_path, content_type=content_type, object_key=object_key)

    async def get_public_url(self, reference: Reference[Any]) -> str:
        """Get a public URL for a reference."""
        object_key = self._object_key_from_reference(reference)
        return self._public_url_for_object(self.bucket, object_key)

corpus_from_minio(*, object_key: str, bucket: Optional[str] = None, content_type: Optional[str] = None, copy_to_task: bool = False, task_id: Optional[str] = None) -> Corpus[bytes] async

Create a corpus pointing to an existing S3-compatible object.

By default this does not copy data into {prefix}/{task_id}. Set copy_to_task=True to materialize a task-scoped duplicate.

Source code in lingo/bucket/archive.py
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
async def corpus_from_minio(
    self,
    *,
    object_key: str,
    bucket: Optional[str] = None,
    content_type: Optional[str] = None,
    copy_to_task: bool = False,
    task_id: Optional[str] = None,
) -> Corpus[bytes]:
    """Create a corpus pointing to an existing S3-compatible object.

    By default this does not copy data into `{prefix}/{task_id}`. Set
    `copy_to_task=True` to materialize a task-scoped duplicate.
    """
    source_bucket = bucket or self.bucket
    source_key = object_key.lstrip("/")

    if not copy_to_task:
        return Corpus[bytes](
            content_type=content_type,
            source_url=self._public_url_for_object(source_bucket, source_key),
        )

    if not task_id:
        raise ValueError("task_id is required when copy_to_task=True")
    if self._s3_client is None:
        raise RuntimeError("copy_to_task requires an initialized S3 client")

    target_ref = await self.create_task_reference(
        task_id=task_id,
        content_type=content_type,
        path_hint=Path(source_key).name,
    )
    target_key = target_ref.object_key or self._object_key_from_reference(target_ref)

    try:
        self._s3_client.copy_object(
            Bucket=self.bucket,
            Key=target_key,
            CopySource={"Bucket": source_bucket, "Key": source_key},
        )
    except Exception:
        source_url = self._public_url_for_object(source_bucket, source_key)
        with urlopen(source_url, timeout=20) as response:
            data = response.read()
        target_ref.dump_content(data, content_type=content_type)
        await self.ensure_reference_uploaded(target_ref)

    return Corpus[bytes](
        content_type=content_type,
        source_url=await self.get_public_url(target_ref),
    )

corpus_from_url(url: str, content_type: Optional[str] = None) -> Corpus[bytes] async

Load a corpus from a URL.

Source code in lingo/bucket/archive.py
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
async def corpus_from_url(
    self,
    url: str,
    content_type: Optional[str] = None,
) -> Corpus[bytes]:
    """Load a corpus from a URL."""
    local_candidate = Path(url)
    if local_candidate.exists():
        return Corpus.from_file(local_candidate, content_type=content_type)

    parsed = urlparse(url)
    if parsed.scheme in ("", "file"):
        local_path = Path(parsed.path if parsed.scheme == "file" else url)
        return Corpus.from_file(local_path, content_type=content_type)

    locator = self._parse_minio_locator(url)
    if locator is not None:
        bucket, object_key = locator
        return await self.corpus_from_minio(
            bucket=bucket,
            object_key=object_key,
            content_type=content_type,
            copy_to_task=False,
        )

    with urlopen(url, timeout=10) as response:
        data = response.read()
    return Corpus[bytes](content_type=content_type, bytes_data=data)

ensure_reference_uploaded(reference: Reference[Any]) -> None async

Upload local reference file to S3-compatible storage when configured.

Source code in lingo/bucket/archive.py
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
async def ensure_reference_uploaded(self, reference: Reference[Any]) -> None:
    """Upload local reference file to S3-compatible storage when configured."""
    if self._s3_client is None:
        return

    local_path = Path(reference.path)
    if not local_path.exists():
        return

    object_key = self._object_key_from_reference(reference)
    extra_args = {"ContentType": reference.content_type or "application/octet-stream"}
    self._s3_client.upload_file(
        str(local_path),
        self.bucket,
        object_key,
        ExtraArgs=extra_args,
    )

get_public_url(reference: Reference[Any]) -> str async

Get a public URL for a reference.

Source code in lingo/bucket/archive.py
307
308
309
310
async def get_public_url(self, reference: Reference[Any]) -> str:
    """Get a public URL for a reference."""
    object_key = self._object_key_from_reference(reference)
    return self._public_url_for_object(self.bucket, object_key)

get_reference(path: Optional[str] = None, content_type: Optional[str] = None, *, task_id: Optional[str] = None) -> Reference[bytes] async

Get a reference to a storage location.

Source code in lingo/bucket/archive.py
295
296
297
298
299
300
301
302
303
304
305
async def get_reference(
    self,
    path: Optional[str] = None,
    content_type: Optional[str] = None,
    *,
    task_id: Optional[str] = None,
) -> Reference[bytes]:
    """Get a reference to a storage location."""
    object_key = self._build_object_key(path=path, task_id=task_id, content_type=content_type)
    full_path = self._local_path_from_object_key(object_key)
    return Reference.from_path(full_path, content_type=content_type, object_key=object_key)

prepare_corpus_for_dispatch(corpus: Corpus[Any], *, task_id: str) -> Corpus[Any] async

Ensure local corpus payload is remotely readable for distributed dispatch.

Source code in lingo/bucket/archive.py
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
async def prepare_corpus_for_dispatch(self, corpus: Corpus[Any], *, task_id: str) -> Corpus[Any]:
    """Ensure local corpus payload is remotely readable for distributed dispatch."""
    source_path = corpus.source_path
    if not source_path:
        return corpus

    local_path = Path(source_path)
    if not local_path.exists():
        return corpus

    reference = await self.create_task_reference(
        task_id=task_id,
        content_type=corpus.content_type,
        path_hint=local_path.name,
    )
    reference.dump_file(local_path, content_type=corpus.content_type)
    await self.ensure_reference_uploaded(reference)
    public_url = await self.get_public_url(reference)
    return Corpus[Any](
        content_type=corpus.content_type,
        source_url=public_url,
        object_data=corpus.object_data,
        bytes_data=None,
        source_path=None,
    )