lumera 0.10.0__py3-none-any.whl → 0.10.2__py3-none-any.whl

lumera/__init__.py

@@ -13,7 +13,7 @@ except PackageNotFoundError:
     __version__ = "unknown"  # Not installed (e.g., running from source)
 
 # Import new modules (as modules, not individual functions)
-from . import automations, email, exceptions, integrations, llm, locks, pb, storage, webhooks
+from . import automations, documents, email, exceptions, integrations, llm, locks, pb, storage, webhooks
 from ._utils import (
     LumeraAPIError,
     RecordNotUniqueError,
@@ -102,6 +102,7 @@ __all__ = [
     "LockHeldError",
     # New modules (use as lumera.pb, lumera.storage, etc.)
     "automations",
+    "documents",
     "email",
     "pb",
     "storage",

lumera/automations.py

@@ -64,6 +64,7 @@ __all__ = [
     # Classes
     "Run",
     "Automation",
+    "LogEntry",
     "LogsResponse",
 ]
 
@@ -71,6 +72,67 @@ from ._utils import LumeraAPIError, _api_request
 from .sdk import get_automation_run as _get_automation_run
 from .sdk import run_automation as _run_automation
 
+# ============================================================================
+# LogEntry Class
+# ============================================================================
+
+
+class LogEntry:
+    """A single log entry from an automation run.
+
+    Attributes:
+        content: The log message content.
+        type: Log type ("stream_stdout", "stream_stderr", "warning", "image_png", "image_jpeg").
+        timestamp: ISO timestamp when the log was emitted.
+        error: True if this is an error entry.
+    """
+
+    def __init__(self, data: dict[str, Any]) -> None:
+        self._data = data
+
+    @property
+    def content(self) -> str:
+        return self._data.get("content", "")
+
+    @property
+    def type(self) -> str:
+        return self._data.get("type", "")
+
+    @property
+    def timestamp(self) -> str | None:
+        return self._data.get("timestamp")
+
+    @property
+    def error(self) -> bool:
+        return self._data.get("error", False)
+
+    @property
+    def is_image(self) -> bool:
+        """True if this entry contains image data (base64 encoded in content)."""
+        return self.type in ("image_png", "image_jpeg")
+
+    def __repr__(self) -> str:
+        preview = self.content[:50] + "..." if len(self.content) > 50 else self.content
+        return f"LogEntry(type={self.type!r}, content={preview!r})"
+
+
+def _parse_ndjson_entries(data: str) -> list[LogEntry]:
+    """Parse NDJSON log data into LogEntry objects."""
+    entries = []
+    for line in data.splitlines():
+        line = line.strip()
+        if not line:
+            continue
+        try:
+            parsed = json.loads(line)
+            if isinstance(parsed, dict):
+                entries.append(LogEntry(parsed))
+        except (json.JSONDecodeError, ValueError):
+            # Skip malformed lines or lines with huge numbers
+            pass
+    return entries
+
+
 # ============================================================================
 # LogsResponse Class
 # ============================================================================
@@ -81,6 +143,7 @@ class LogsResponse:
 
     Attributes:
         data: Raw log content as a string (NDJSON format).
+        entries: Parsed log entries as LogEntry objects.
         offset: Byte offset where this chunk starts.
         size: Number of bytes in this chunk.
         total_size: Total size of the log file.
@@ -91,11 +154,20 @@ class LogsResponse:
 
     def __init__(self, data: dict[str, Any]) -> None:
         self._data = data
+        self._entries: list[LogEntry] | None = None
 
     @property
     def data(self) -> str:
+        """Raw NDJSON log content."""
         return self._data.get("data", "")
 
+    @property
+    def entries(self) -> list[LogEntry]:
+        """Parsed log entries. Lazily parsed from NDJSON data."""
+        if self._entries is None:
+            self._entries = _parse_ndjson_entries(self.data)
+        return self._entries
+
     @property
     def offset(self) -> int:
         return self._data.get("offset", 0)
@@ -342,7 +414,7 @@ class Run:
             raise ValueError("Cannot fetch logs without run id")
         return get_logs(self.id, offset=offset, limit=limit, all=all)
 
-    def stream_logs(self, *, timeout: float = 30) -> Iterator[str]:
+    def stream_logs(self, *, timeout: float = 30) -> Iterator[LogEntry]:
         """Stream logs from this run.
 
         Works for both live (running) and archived (completed) runs.
@@ -353,12 +425,12 @@ class Run:
             timeout: HTTP connection timeout in seconds.
 
         Yields:
-            Log lines as strings (raw NDJSON lines).
+            LogEntry objects with content, type, timestamp, and error fields.
 
         Example:
             >>> run = automations.run("automation_id", inputs={})
-            >>> for line in run.stream_logs():
-            ...     print(line)
+            >>> for entry in run.stream_logs():
+            ...     print(f"[{entry.type}] {entry.content}")
         """
         if not self.id:
             raise ValueError("Cannot stream logs without run id")
@@ -912,11 +984,11 @@ def delete(automation_id: str) -> None:
 # ============================================================================
 
 
-def stream_logs(run_id: str, *, timeout: float = 30) -> Iterator[str]:
+def stream_logs(run_id: str, *, timeout: float = 30) -> Iterator[LogEntry]:
     """Stream logs from an automation run.
 
     Works for both live (running) and archived (completed) runs.
-    Connects to the server-sent events endpoint and yields log lines
+    Connects to the server-sent events endpoint and yields LogEntry objects
     as they arrive. For live runs, streams in real-time. For archived
     runs, streams the entire log from storage.
 
@@ -925,11 +997,11 @@ def stream_logs(run_id: str, *, timeout: float = 30) -> Iterator[str]:
         timeout: HTTP connection timeout in seconds.
 
     Yields:
-        Log lines as strings (raw NDJSON lines).
+        LogEntry objects with content, type, timestamp, and error fields.
 
     Example:
-        >>> for line in automations.stream_logs("run_id"):
-        ...     print(line)
+        >>> for entry in automations.stream_logs("run_id"):
+        ...     print(f"[{entry.type}] {entry.content}")
     """
     import base64
     import os
@@ -971,10 +1043,20 @@ def stream_logs(run_id: str, *, timeout: float = 30) -> Iterator[str]:
                     try:
                         data = json.loads(current_data)
                         if "data" in data:
-                            # Data is base64-encoded
+                            # Data is base64-encoded NDJSON
                             raw = base64.b64decode(data["data"])
                             decoded = raw.decode("utf-8", errors="replace")
-                            yield from decoded.splitlines()
+                            for ndjson_line in decoded.splitlines():
+                                ndjson_line = ndjson_line.strip()
+                                if not ndjson_line:
+                                    continue
+                                try:
+                                    entry_data = json.loads(ndjson_line)
+                                    if isinstance(entry_data, dict):
+                                        yield LogEntry(entry_data)
+                                except (json.JSONDecodeError, ValueError):
+                                    # Skip malformed lines or lines with huge numbers
+                                    pass
                     except (json.JSONDecodeError, KeyError):
                         pass
                 elif current_event == "complete":

lumera/documents.py

@@ -0,0 +1,127 @@
+"""
+Document text extraction via OpenAI Responses API.
+
+Supports images and PDFs.
+
+Example:
+    >>> from lumera import documents
+    >>> text = documents.extract_text("invoice.pdf")
+
+    # Or from bytes (no file needed):
+    >>> text = documents.extract_text_from_bytes(pdf_bytes, "application/pdf")
+"""
+
+from __future__ import annotations
+
+import base64
+import mimetypes
+from pathlib import Path
+
+__all__ = ["extract_text", "extract_text_from_bytes"]
+
+_DEFAULT_MODEL = "gpt-5-mini"
+_DEFAULT_PROMPT = "Extract all text from this document. Return only the extracted text."
+
+
+def _get_mime_type(file_path: str) -> str:
+    """Get MIME type for a file."""
+    mime, _ = mimetypes.guess_type(file_path)
+    if mime:
+        return mime
+    ext = Path(file_path).suffix.lower()
+    return {
+        ".png": "image/png",
+        ".jpg": "image/jpeg",
+        ".jpeg": "image/jpeg",
+        ".gif": "image/gif",
+        ".webp": "image/webp",
+        ".pdf": "application/pdf",
+    }.get(ext, "application/octet-stream")
+
+
+def extract_text_from_bytes(
+    data: bytes,
+    mime_type: str,
+    *,
+    filename: str = "document",
+    prompt: str = _DEFAULT_PROMPT,
+    model: str = _DEFAULT_MODEL,
+) -> str:
+    """Extract text from document bytes using OpenAI.
+
+    Args:
+        data: Raw file bytes
+        mime_type: MIME type (e.g., "application/pdf", "image/png")
+        filename: Optional filename for context
+        prompt: What to ask the LLM
+        model: Model to use (default: gpt-5-mini)
+
+    Returns:
+        Extracted text
+
+    Example:
+        >>> import requests
+        >>> resp = requests.get(presigned_url)
+        >>> text = documents.extract_text_from_bytes(
+        ...     resp.content,
+        ...     "application/pdf",
+        ...     filename="invoice.pdf"
+        ... )
+    """
+    from . import llm
+
+    b64 = base64.b64encode(data).decode("utf-8")
+    data_url = f"data:{mime_type};base64,{b64}"
+    client = llm.get_provider().client
+
+    response = client.responses.create(
+        model=model,
+        input=[
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "input_file",
+                        "filename": filename,
+                        "file_data": data_url,
+                    },
+                    {
+                        "type": "input_text",
+                        "text": prompt,
+                    },
+                ],
+            },
+        ],
+    )
+    return response.output_text or ""
+
+
+def extract_text(
+    file_path: str,
+    *,
+    prompt: str = _DEFAULT_PROMPT,
+    model: str = _DEFAULT_MODEL,
+) -> str:
+    """Extract text from a document file using OpenAI.
+
+    Args:
+        file_path: Path to image or PDF
+        prompt: What to ask the LLM
+        model: Model to use (default: gpt-5-mini)
+
+    Returns:
+        Extracted text
+    """
+    with open(file_path, "rb") as f:
+        data = f.read()
+
+    mime = _get_mime_type(file_path)
+    filename = Path(file_path).name
+
+    return extract_text_from_bytes(
+        data,
+        mime,
+        filename=filename,
+        prompt=prompt,
+        model=model,
+    )

lumera/llm.py

@@ -16,7 +16,7 @@ Configuration:
 
 Example:
     >>> from lumera import llm
-    >>> response = llm.complete("What is 2+2?", model="gpt-5.2-mini")
+    >>> response = llm.complete("What is 2+2?", model="gpt-5-mini")
     >>> print(response["content"])
 """
 
@@ -135,14 +135,14 @@ class OpenAIProvider(LLMProvider):
     # Model aliases for convenience
     MODEL_ALIASES: dict[str, str] = {
         "gpt-5.2": "gpt-5.2",
-        "gpt-5.2-mini": "gpt-5.2-mini",
-        "gpt-5.2-nano": "gpt-5.2-nano",
+        "gpt-5-mini": "gpt-5-mini",
+        "gpt-5-nano": "gpt-5-nano",
         # Embedding models
         "text-embedding-3-small": "text-embedding-3-small",
         "text-embedding-3-large": "text-embedding-3-large",
     }
 
-    DEFAULT_CHAT_MODEL = "gpt-5.2-mini"
+    DEFAULT_CHAT_MODEL = "gpt-5-mini"
     DEFAULT_EMBEDDING_MODEL = "text-embedding-3-small"
     DEFAULT_PROVIDER_NAME = "openai"
 
@@ -374,7 +374,7 @@ def set_provider(provider: LLMProvider | str, **kwargs: Unpack[ProviderConfig])
 def complete(
     prompt: str,
     *,
-    model: str = "gpt-5.2-mini",
+    model: str = "gpt-5-mini",
     temperature: float = 0.7,
     max_tokens: int | None = None,
     system_prompt: str | None = None,
@@ -384,7 +384,7 @@ def complete(
 
     Args:
         prompt: User prompt/question
-        model: Model to use (default: gpt-5.2-mini)
+        model: Model to use (default: gpt-5-mini)
         temperature: Sampling temperature 0.0 to 2.0 (default: 0.7)
         max_tokens: Max tokens in response (None = model default)
         system_prompt: Optional system message to set behavior
@@ -397,7 +397,7 @@ def complete(
         >>> response = llm.complete(
         ...     prompt="Classify this deposit: ...",
         ...     system_prompt="You are an expert accountant.",
-        ...     model="gpt-5.2-mini",
+        ...     model="gpt-5-mini",
         ...     json_mode=True
         ... )
         >>> data = json.loads(response["content"])
@@ -416,7 +416,7 @@ def complete(
 def chat(
     messages: list[Message],
     *,
-    model: str = "gpt-5.2-mini",
+    model: str = "gpt-5-mini",
     temperature: float = 0.7,
     max_tokens: int | None = None,
     json_mode: bool = False,
@@ -425,7 +425,7 @@ def chat(
 
     Args:
         messages: Conversation history with role and content
-        model: Model to use (default: gpt-5.2-mini)
+        model: Model to use (default: gpt-5-mini)
         temperature: Sampling temperature 0.0 to 2.0 (default: 0.7)
         max_tokens: Max tokens in response (None = model default)
         json_mode: Force JSON output (default: False)

{lumera-0.10.0.dist-info → lumera-0.10.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lumera
-Version: 0.10.0
+Version: 0.10.2
 Summary: SDK for building on Lumera platform
 Requires-Python: >=3.11
 Requires-Dist: requests

{lumera-0.10.0.dist-info → lumera-0.10.2.dist-info}/RECORD

@@ -1,11 +1,12 @@
-lumera/__init__.py,sha256=5FlY5dSJ1WNM4ko7wgmcajO8G2voBGn4S19E91_WdqE,2687
+lumera/__init__.py,sha256=TraEFeZvI92sDE88amgTctLzAapnedtOijoY4XDRwVE,2715
 lumera/_utils.py,sha256=b-l3Ebh4n2pC-9T5mR6h4hPf_Wl48VDlHES0pLo1zKE,25766
-lumera/automations.py,sha256=KPP_rD7WKmBs865jiKoonZJjdTno-FSAU7hajPFyqs0,32851
+lumera/automations.py,sha256=NhUKizNpw0p_EQVpsrLCfmZ7c0Xmh5XOggKdoWJxam0,35869
+lumera/documents.py,sha256=n585BOhVntvhh6v9uip4Cug4c9N-wu-rTNFLsNHWN6Y,3156
 lumera/email.py,sha256=lk8KUsRw1ZvxgM0FPQXH-jVKUQA5f0zLv88jlc3IWlA,5056
 lumera/exceptions.py,sha256=bNsx4iYaroAAGsYxErfELC2B5ZJ3w5lVa1kKdIx5s9g,2173
 lumera/files.py,sha256=xMJmLTSaQQDttM3AMmpOWc6soh4lvCCKBreV0fXWHQw,3159
 lumera/google.py,sha256=zpWW1qSlzLZY5Ip7cGAzrv9sJrQf3JBKH2ODc1cCM_E,1130
-lumera/llm.py,sha256=pUTZK7t3GTK0vfxMI1PJgJwNendyuiJc5MB1pUj2vxE,14412
+lumera/llm.py,sha256=HuZ-IUbszCCE0vAta9k89XvKt8_F9qVgF201CoKyYYI,14390
 lumera/locks.py,sha256=8l_qxb8nrxge7YJ-ApUTJ5MeYpIdxDeEa94Eim9O-YM,6806
 lumera/pb.py,sha256=Q_U1cKeB3YgI7bmTquzLYFWTRWcfUZkFSl7JXMBzV7M,20700
 lumera/sdk.py,sha256=Dw0yxlZ-ncjgPkCpVnAJQIURtIsbUA4RVu9VjXLayDc,34078
@@ -13,7 +14,7 @@ lumera/storage.py,sha256=fWkscTvKDzQ-5tsfA1lREO2qgtjJ4Yvxj3hvYNLKiW0,10527
 lumera/webhooks.py,sha256=L_Q5YHBJKQNpv7G9Nq0QqlGMRch6x9ptlwu1xD2qwUc,8661
 lumera/integrations/__init__.py,sha256=LnJmAnFB_p3YMKyeGVdDP4LYlJ85XFNQFAxGo6zF7CI,937
 lumera/integrations/google.py,sha256=QkbBbbDh3I_OToPDFqcivU6sWy2UieHBxZ_TPv5rqK0,11862
-lumera-0.10.0.dist-info/METADATA,sha256=uWvSDuD868zVICFyVUppHMIrWe6A-JKeyxFRurjxieU,1612
-lumera-0.10.0.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
-lumera-0.10.0.dist-info/top_level.txt,sha256=HgfK4XQkpMTnM2E5iWM4kB711FnYqUY9dglzib3pWlE,7
-lumera-0.10.0.dist-info/RECORD,,
+lumera-0.10.2.dist-info/METADATA,sha256=zQEAmPgFin8QKsWcsLQN3QpqvPRndihzwh6s8E9J8Y4,1612
+lumera-0.10.2.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+lumera-0.10.2.dist-info/top_level.txt,sha256=HgfK4XQkpMTnM2E5iWM4kB711FnYqUY9dglzib3pWlE,7
+lumera-0.10.2.dist-info/RECORD,,