vespera 0.1.0__py3-none-any.whl

vespera/__init__.py

@@ -0,0 +1,3 @@
+"""Vespera: local-first AI due diligence."""
+
+__version__ = "0.1.0"

vespera/cli.py

@@ -0,0 +1,133 @@
+"""Vespera command-line interface."""
+
+from collections import Counter
+from pathlib import Path
+
+import typer
+from rich.console import Console
+from rich.progress import BarColumn, Progress, TaskProgressColumn, TextColumn, TimeElapsedColumn
+
+import vespera
+from vespera.config import DEFAULT_MODEL, DEFAULT_OLLAMA_HOST, ReviewConfig
+from vespera.documents.loader import discover_documents, load_document
+from vespera.llm.ollama import OllamaError, OllamaProvider
+from vespera.review.aggregator import aggregate_findings
+from vespera.review.analyzer import analyze_document, cross_document_findings
+from vespera.review.models import DocumentSummary, Finding
+from vespera.review.report import write_outputs
+
+app = typer.Typer(add_completion=False, no_args_is_help=True)
+console = Console()
+
+
+def _version_callback(value: bool):
+    if value:
+        console.print(f"vespera {vespera.__version__}")
+        raise typer.Exit()
+
+
+@app.callback()
+def main(
+    version: bool = typer.Option(
+        False, "--version", callback=_version_callback, is_eager=True, help="Show version."
+    ),
+):
+    """Vespera — local-first AI due diligence."""
+
+
+@app.command()
+def review(
+    path: Path = typer.Argument(..., exists=True, file_okay=False, help="Dataroom directory."),
+    model: str = typer.Option(DEFAULT_MODEL, "--model", "-m", help="Ollama model to use."),
+    output: Path = typer.Option(Path("vespera-output"), "--output", "-o", help="Output directory."),
+    host: str = typer.Option(DEFAULT_OLLAMA_HOST, "--host", help="Ollama server URL."),
+):
+    """Review a local dataroom and produce a due diligence report."""
+    config = ReviewConfig(model=model, ollama_host=host, output_dir=output)
+    provider = OllamaProvider(model=config.model, host=config.ollama_host)
+
+    console.print("\n[bold]Vespera[/bold]\n")
+    console.print(f"Reviewing [cyan]{path}[/cyan]\n")
+
+    paths = discover_documents(path)
+    console.print(f"Documents found: {len(paths)}")
+    if not paths:
+        console.print("[yellow]No supported documents (.pdf, .docx, .txt, .md) found.[/yellow]")
+        raise typer.Exit(code=1)
+
+    findings: list[Finding] = []
+    summaries: dict[str, DocumentSummary] = {}
+    reviewed: list[str] = []
+    empty: list[str] = []
+    processed = 0
+
+    try:
+        with Progress(
+            TextColumn("[progress.description]{task.description}"),
+            BarColumn(),
+            TaskProgressColumn(),
+            TimeElapsedColumn(),
+            console=console,
+        ) as progress:
+            task = progress.add_task("Analysing documents", total=len(paths))
+            for doc_path in paths:
+                relative_name = str(doc_path.relative_to(path))
+                progress.update(task, description=f"Analysing {relative_name}")
+                document = load_document(doc_path)
+                if document.is_empty:
+                    empty.append(relative_name)
+                else:
+                    doc_findings, summary = analyze_document(
+                        document, provider, config, relative_name
+                    )
+                    findings.extend(doc_findings)
+                    if summary is not None:
+                        summaries[relative_name] = summary
+                    reviewed.append(relative_name)
+                processed += 1
+                progress.advance(task)
+
+            cross_task = progress.add_task("Cross-referencing documents", total=1)
+            findings.extend(cross_document_findings(summaries, provider, config))
+            progress.advance(cross_task)
+    except OllamaError as error:
+        console.print(f"\n[red]Error:[/red] {error}")
+        raise typer.Exit(code=1)
+
+    findings = aggregate_findings(findings)
+    report_path, findings_path = write_outputs(findings, reviewed, config.output_dir, empty)
+
+    console.print(f"Documents processed: {processed}\n")
+    console.print("[bold]Findings:[/bold]")
+    category_counts = Counter(f.category for f in findings)
+    if category_counts:
+        for category, count in category_counts.most_common():
+            console.print(f"- {category[0].upper() + category[1:]}: {count}")
+    else:
+        console.print("- No findings recorded")
+    console.print(f"\nReport: [green]{report_path}[/green]")
+    console.print(f"Evidence: [green]{findings_path}[/green]")
+    console.print("\n[dim]All document analysis was performed locally.[/dim]\n")
+
+
+@app.command()
+def models(
+    host: str = typer.Option(DEFAULT_OLLAMA_HOST, "--host", help="Ollama server URL."),
+):
+    """Show the default model and locally installed Ollama models."""
+    console.print(f"Default model: [cyan]{DEFAULT_MODEL}[/cyan]")
+    provider = OllamaProvider(model=DEFAULT_MODEL, host=host)
+    local = provider.list_local_models()
+    if local:
+        console.print("Locally installed Ollama models:")
+        for name in local:
+            marker = " [green](default)[/green]" if name == DEFAULT_MODEL else ""
+            console.print(f"- {name}{marker}")
+    else:
+        console.print(
+            f"[yellow]Could not list local models — is Ollama running at {host}?[/yellow]"
+        )
+
+
+if __name__ == "__main__":
+    app()

vespera/config.py

@@ -0,0 +1,18 @@
+"""Configuration defaults for a review run."""
+
+import os
+from dataclasses import dataclass, field
+from pathlib import Path
+
+DEFAULT_MODEL = "qwen3:8b"
+DEFAULT_OLLAMA_HOST = os.environ.get("VESPERA_OLLAMA_HOST", "http://localhost:11434")
+
+
+@dataclass
+class ReviewConfig:
+    model: str = DEFAULT_MODEL
+    ollama_host: str = DEFAULT_OLLAMA_HOST
+    chunk_chars: int = 7000
+    chunk_overlap_chars: int = 400
+    max_evidence_chars: int = 300
+    output_dir: Path = field(default_factory=lambda: Path("vespera-output"))

vespera/documents/docx_.py

@@ -0,0 +1,14 @@
+"""DOCX text extraction via python-docx."""
+
+from pathlib import Path
+
+import docx
+
+
+def extract_text(path: Path) -> str:
+    document = docx.Document(str(path))
+    parts = [paragraph.text for paragraph in document.paragraphs]
+    for table in document.tables:
+        for row in table.rows:
+            parts.append(" | ".join(cell.text for cell in row.cells))
+    return "\n".join(parts)

vespera/documents/loader.py

@@ -0,0 +1,52 @@
+"""Document discovery and text extraction."""
+
+from dataclasses import dataclass
+from pathlib import Path
+
+from vespera.documents import docx_, pdf
+
+SUPPORTED_EXTENSIONS = {".pdf", ".docx", ".txt", ".md"}
+
+
+@dataclass
+class Page:
+    number: int | None  # 1-based page number, None when the format has no pages
+    text: str
+
+
+@dataclass
+class Document:
+    path: Path
+    pages: list[Page]
+
+    @property
+    def text(self) -> str:
+        return "\n".join(page.text for page in self.pages)
+
+    @property
+    def is_empty(self) -> bool:
+        return not self.text.strip()
+
+
+def discover_documents(root: Path) -> list[Path]:
+    """Recursively find supported documents under root, skipping hidden files/dirs."""
+    found = []
+    for path in sorted(root.rglob("*")):
+        if not path.is_file():
+            continue
+        if any(part.startswith(".") for part in path.relative_to(root).parts):
+            continue
+        if path.suffix.lower() in SUPPORTED_EXTENSIONS:
+            found.append(path)
+    return found
+
+
+def load_document(path: Path) -> Document:
+    suffix = path.suffix.lower()
+    if suffix == ".pdf":
+        pages = pdf.extract_pages(path)
+    elif suffix == ".docx":
+        pages = [Page(number=None, text=docx_.extract_text(path))]
+    else:  # .txt, .md
+        pages = [Page(number=None, text=path.read_text(encoding="utf-8", errors="replace"))]
+    return Document(path=path, pages=pages)

vespera/documents/pdf.py

@@ -0,0 +1,15 @@
+"""PDF text extraction via PyMuPDF."""
+
+from pathlib import Path
+
+import pymupdf
+
+
+def extract_pages(path: Path) -> list:
+    from vespera.documents.loader import Page
+
+    pages = []
+    with pymupdf.open(path) as doc:
+        for index, page in enumerate(doc, start=1):
+            pages.append(Page(number=index, text=page.get_text()))
+    return pages

vespera/llm/base.py

@@ -0,0 +1,9 @@
+"""Minimal LLM provider interface so the backend stays replaceable."""
+
+from typing import Protocol
+
+from pydantic import BaseModel
+
+
+class LLMProvider(Protocol):
+    def generate_structured(self, prompt: str, schema: type[BaseModel]) -> BaseModel: ...

vespera/llm/ollama.py

@@ -0,0 +1,59 @@
+"""Ollama-backed LLM provider using the local HTTP API."""
+
+import httpx
+from pydantic import BaseModel, ValidationError
+
+
+class OllamaError(RuntimeError):
+    pass
+
+
+class OllamaProvider:
+    def __init__(self, model: str, host: str = "http://localhost:11434", timeout: float = 300.0):
+        self.model = model
+        self.host = host.rstrip("/")
+        self._client = httpx.Client(base_url=self.host, timeout=timeout)
+
+    def generate_structured(self, prompt: str, schema: type[BaseModel]) -> BaseModel:
+        last_error: Exception | None = None
+        for _ in range(2):  # one retry on invalid output
+            content = self._chat(prompt, schema)
+            try:
+                return schema.model_validate_json(content)
+            except ValidationError as error:
+                last_error = error
+        raise OllamaError(f"Model returned output that failed validation twice: {last_error}")
+
+    def _chat(self, prompt: str, schema: type[BaseModel]) -> str:
+        payload = {
+            "model": self.model,
+            "messages": [{"role": "user", "content": prompt}],
+            "stream": False,
+            "format": schema.model_json_schema(),
+            "think": False,
+            "options": {"temperature": 0.1},
+        }
+        try:
+            response = self._client.post("/api/chat", json=payload)
+            response.raise_for_status()
+        except httpx.ConnectError as error:
+            raise OllamaError(
+                f"Cannot reach Ollama at {self.host}. Is it running? "
+                "Install from https://ollama.com and run: ollama serve"
+            ) from error
+        except httpx.HTTPStatusError as error:
+            detail = error.response.text[:300]
+            if error.response.status_code == 404:
+                raise OllamaError(
+                    f"Model '{self.model}' not found. Pull it first: ollama pull {self.model}"
+                ) from error
+            raise OllamaError(f"Ollama request failed: {detail}") from error
+        return response.json()["message"]["content"]
+
+    def list_local_models(self) -> list[str]:
+        try:
+            response = self._client.get("/api/tags")
+            response.raise_for_status()
+        except httpx.HTTPError:
+            return []
+        return [item["name"] for item in response.json().get("models", [])]

vespera/review/aggregator.py

@@ -0,0 +1,45 @@
+"""Aggregate and deduplicate findings across chunks and documents."""
+
+import re
+
+from vespera.review.models import SEVERITY_ORDER, Finding
+
+
+def _normalize(title: str) -> str:
+    return re.sub(r"[^a-z0-9 ]", "", title.lower()).strip()
+
+
+def _similar(a: str, b: str) -> bool:
+    """Cheap fuzzy match: identical normalized titles, or high word overlap."""
+    if a == b:
+        return True
+    words_a, words_b = set(a.split()), set(b.split())
+    if not words_a or not words_b:
+        return False
+    overlap = len(words_a & words_b) / min(len(words_a), len(words_b))
+    return overlap >= 0.8
+
+
+def aggregate_findings(findings: list[Finding]) -> list[Finding]:
+    """Collapse near-duplicate findings (same category + file + similar title).
+
+    Keeps the highest-confidence instance. Result is sorted by severity then confidence.
+    """
+    kept: list[Finding] = []
+    for finding in findings:
+        duplicate_of = None
+        for existing in kept:
+            if (
+                existing.category == finding.category
+                and existing.source_file == finding.source_file
+                and _similar(_normalize(existing.title), _normalize(finding.title))
+            ):
+                duplicate_of = existing
+                break
+        if duplicate_of is None:
+            kept.append(finding)
+        elif finding.confidence > duplicate_of.confidence:
+            kept[kept.index(duplicate_of)] = finding
+
+    kept.sort(key=lambda f: (SEVERITY_ORDER[f.severity], -f.confidence, f.source_file))
+    return kept

vespera/review/analyzer.py

@@ -0,0 +1,145 @@
+"""Chunking and per-document / cross-document analysis."""
+
+import json
+import re
+from dataclasses import dataclass
+
+from vespera.config import ReviewConfig
+from vespera.documents.loader import Document
+from vespera.llm.base import LLMProvider
+from vespera.review import prompts
+from vespera.review.models import (
+    ChunkFindings,
+    CrossDocumentFindings,
+    DocumentSummary,
+    Finding,
+)
+
+
+@dataclass
+class Chunk:
+    text: str
+    start_page: int | None
+
+
+# small local models like to report "signatures present" as a "missing signatures"
+# finding despite instructions; a signature finding that affirms completeness is noise
+_POSITIVE_SIGNATURE_NOTE = re.compile(
+    r"signatures? (are |is )?present"
+    r"|(is|are|been|was|were|parties)( duly| fully)? signed"
+    r"|fully executed|duly executed"
+    r"|signature (blocks?|areas?) (is|are) complete"
+    r"|(both|all) parties (have |has )?signed",
+    re.IGNORECASE,
+)
+
+
+def _is_noise(finding) -> bool:
+    if finding.category != "missing signatures":
+        return False
+    if finding.severity == "info":
+        return True
+    return bool(_POSITIVE_SIGNATURE_NOTE.search(f"{finding.title} {finding.summary}"))
+
+
+def chunk_document(document: Document, chunk_chars: int, overlap_chars: int = 0) -> list[Chunk]:
+    """Split a document into chunks, keeping track of the page each chunk starts on."""
+    chunks: list[Chunk] = []
+    current_text = ""
+    current_page: int | None = None
+
+    for page in document.pages:
+        remaining = page.text
+        while remaining:
+            if not current_text:
+                current_page = page.number
+            space = chunk_chars - len(current_text)
+            current_text += remaining[:space]
+            remaining = remaining[space:]
+            if len(current_text) >= chunk_chars:
+                chunks.append(Chunk(text=current_text, start_page=current_page))
+                # carry a small overlap so clauses split across a boundary aren't lost
+                current_text = current_text[-overlap_chars:] if overlap_chars else ""
+    if current_text.strip():
+        chunks.append(Chunk(text=current_text, start_page=current_page))
+    return chunks
+
+
+def analyze_document(
+    document: Document,
+    provider: LLMProvider,
+    config: ReviewConfig,
+    relative_name: str,
+) -> tuple[list[Finding], DocumentSummary | None]:
+    """Extract findings from one document, plus a summary for the cross-document pass."""
+    findings: list[Finding] = []
+    for chunk in chunk_document(document, config.chunk_chars, config.chunk_overlap_chars):
+        page_info = f"Page: {chunk.start_page}" if chunk.start_page else "Page: not applicable"
+        prompt = prompts.CHUNK_PROMPT.format(
+            role=prompts.ANALYST_ROLE,
+            source_file=relative_name,
+            page_info=page_info,
+            chunk_text=chunk.text,
+        )
+        result = provider.generate_structured(prompt, ChunkFindings)
+        for extracted in result.findings:
+            if _is_noise(extracted):
+                continue
+            findings.append(
+                Finding(
+                    category=extracted.category,
+                    title=extracted.title,
+                    summary=extracted.summary,
+                    severity=extracted.severity,
+                    source_file=relative_name,
+                    source_page=chunk.start_page,
+                    evidence=extracted.evidence[: config.max_evidence_chars],
+                    confidence=extracted.confidence,
+                )
+            )
+
+    summary: DocumentSummary | None = None
+    summary_prompt = prompts.SUMMARY_PROMPT.format(
+        role=prompts.ANALYST_ROLE,
+        source_file=relative_name,
+        document_text=document.text[: config.chunk_chars],
+    )
+    try:
+        summary = provider.generate_structured(summary_prompt, DocumentSummary)
+    except Exception:
+        summary = None  # the cross-document pass is best-effort
+    return findings, summary
+
+
+def cross_document_findings(
+    summaries: dict[str, DocumentSummary],
+    provider: LLMProvider,
+    config: ReviewConfig,
+) -> list[Finding]:
+    """One pass over all document summaries to spot missing references and conflicts."""
+    if len(summaries) < 2:
+        return []
+    file_list = "\n".join(f"- {name}" for name in summaries)
+    rendered = "\n".join(
+        f"### {name}\n{json.dumps(summary.model_dump(), indent=2)}"
+        for name, summary in summaries.items()
+    )
+    prompt = prompts.CROSS_DOCUMENT_PROMPT.format(
+        role=prompts.ANALYST_ROLE, file_list=file_list, summaries=rendered
+    )
+    result = provider.generate_structured(prompt, CrossDocumentFindings)
+    allowed = {"missing documents explicitly referenced elsewhere", "inconsistencies between documents"}
+    return [
+        Finding(
+            category=extracted.category,
+            title=extracted.title,
+            summary=extracted.summary,
+            severity=extracted.severity,
+            source_file="(multiple documents)",
+            source_page=None,
+            evidence=extracted.evidence[: config.max_evidence_chars],
+            confidence=extracted.confidence,
+        )
+        for extracted in result.findings
+        if extracted.category in allowed
+    ]

vespera/review/models.py

@@ -0,0 +1,91 @@
+"""Pydantic models for structured findings."""
+
+from typing import Literal, get_args
+
+from pydantic import BaseModel, Field, field_validator
+
+Category = Literal[
+    "parties",
+    "important dates",
+    "contract type",
+    "governing law",
+    "change-of-control clauses",
+    "termination rights",
+    "assignment restrictions",
+    "exclusivity",
+    "unusual obligations",
+    "material liabilities",
+    "IP ownership / assignment",
+    "confidentiality obligations",
+    "missing signatures",
+    "missing documents explicitly referenced elsewhere",
+    "inconsistencies between documents",
+    "potential red flags",
+]
+
+CATEGORY_ORDER: list[str] = list(get_args(Category))
+
+Severity = Literal["info", "low", "medium", "high"]
+
+SEVERITY_ORDER = {"high": 0, "medium": 1, "low": 2, "info": 3}
+
+
+class Finding(BaseModel):
+    category: str
+    title: str
+    summary: str
+    severity: Severity
+    source_file: str
+    source_page: int | None
+    evidence: str
+    confidence: float = Field(ge=0.0, le=1.0)
+
+
+class ExtractedFinding(BaseModel):
+    """What the LLM returns for a chunk; source location is stamped by code."""
+
+    category: Category
+    title: str
+    summary: str
+    severity: Severity
+    evidence: str = Field(description="Short verbatim excerpt from the text, max ~40 words")
+    confidence: float = Field(ge=0.0, le=1.0, description="Between 0.0 and 1.0")
+
+    @field_validator("confidence", mode="before")
+    @classmethod
+    def _coerce_percentage(cls, value):
+        # local models often answer 90 when they mean 0.9
+        if isinstance(value, (int, float)) and 1.0 < value <= 100.0:
+            return value / 100.0
+        return value
+
+
+class ChunkFindings(BaseModel):
+    findings: list[ExtractedFinding] = Field(
+        default_factory=list,
+        description="Findings supported by the supplied text. Empty if none.",
+    )
+
+
+class DocumentSummary(BaseModel):
+    """One-line facts per document, used for the cross-document pass."""
+
+    contract_type: str = Field(description="e.g. 'Master Services Agreement', or 'unknown'")
+    parties: list[str] = Field(default_factory=list)
+    key_dates: list[str] = Field(default_factory=list)
+    referenced_documents: list[str] = Field(
+        default_factory=list,
+        description="Other documents, schedules, or exhibits this document explicitly refers to",
+    )
+    key_facts: list[str] = Field(
+        default_factory=list,
+        description=(
+            "Material factual assertions worth cross-checking: monetary amounts and "
+            "commitments, ownership claims (e.g. who owns which IP), and key terms"
+        ),
+    )
+    signed: bool = Field(description="Whether the document appears to be executed/signed")
+
+
+class CrossDocumentFindings(BaseModel):
+    findings: list[ExtractedFinding] = Field(default_factory=list)

vespera/review/prompts.py

@@ -0,0 +1,83 @@
+"""Prompts for the due diligence analysis passes."""
+
+ANALYST_ROLE = """\
+You are a cautious due diligence analyst performing a first-pass review of documents \
+in a dataroom for an M&A / investment transaction. You are not a lawyer and you never \
+give legal advice or definitive legal conclusions. You only report findings that are \
+directly supported by the supplied text. You distinguish facts from possible concerns. \
+If the text does not support any finding, return an empty findings list — never invent \
+or speculate beyond the text.\
+"""
+
+CHUNK_PROMPT = """\
+{role}
+
+Review the following excerpt from a document and extract due diligence findings.
+
+Document: {source_file}
+{page_info}
+
+Rules:
+- Only report what this text supports. No speculation. Empty list is a valid answer.
+- "evidence" must be a short verbatim quote from the text (under 40 words).
+- "confidence" is a number between 0.0 and 1.0 reflecting how clearly the text
+  supports the finding.
+- Severity guide: "info" for neutral facts (parties, dates, contract type, governing law);
+  "low"/"medium" for terms a buyer should review (termination rights, assignment
+  restrictions, confidentiality); "medium"/"high" for terms that can materially affect
+  a transaction (change-of-control, exclusivity, uncapped liabilities, missing IP
+  assignment). A "missing signatures" finding is always at least "medium".
+- Report a "missing signatures" finding ONLY when a signature area is visibly blank or
+  incomplete. Never report a finding to say that signatures ARE present, and never
+  report "missing signatures" merely because this excerpt lacks a signature block.
+- Never report a finding whose only content is that something is in order or
+  unremarkable.
+- Use the "contract type" category for at most one finding that identifies what kind of
+  document this is. Commercial terms such as fees, minimum purchase commitments, or
+  price adjustment rights belong under "material liabilities", "unusual obligations",
+  or "potential red flags".
+
+--- DOCUMENT TEXT START ---
+{chunk_text}
+--- DOCUMENT TEXT END ---
+"""
+
+SUMMARY_PROMPT = """\
+{role}
+
+Summarise the key facts of this document for cross-referencing against other documents
+in the same dataroom. Only state what the text supports; use "unknown" when unclear.
+For "signed": true only if the text shows completed signatures (names/dates filled in).
+
+Document: {source_file}
+
+--- DOCUMENT TEXT START ---
+{document_text}
+--- DOCUMENT TEXT END ---
+"""
+
+CROSS_DOCUMENT_PROMPT = """\
+{role}
+
+Below are structured summaries of every document found in a dataroom. Compare them and
+report only two kinds of findings:
+
+1. category "missing documents explicitly referenced elsewhere": check every entry in
+   each document's "referenced_documents" against the list of documents present; report
+   any referenced schedule, exhibit, or agreement that is not in the dataroom.
+2. category "inconsistencies between documents": compare the "key_facts" across
+   documents and report conflicting facts — for example different monetary amounts for
+   the same commitment, contradictory claims about who owns an asset or intellectual
+   property, different names for the same party, or conflicting dates.
+
+Rules:
+- Only report what these summaries support. Empty list is a valid answer.
+- "evidence" must name the documents involved and the conflicting or missing item.
+- Do not repeat single-document findings (signatures, clauses); those are handled elsewhere.
+
+Documents present in the dataroom:
+{file_list}
+
+Document summaries:
+{summaries}
+"""

vespera/review/report.py

@@ -0,0 +1,108 @@
+"""Render the Markdown report and findings JSON."""
+
+import json
+from datetime import date
+from pathlib import Path
+
+from vespera.review.models import CATEGORY_ORDER, SEVERITY_ORDER, Finding
+
+DISCLAIMER = (
+    "> **Important:** This report was produced by automated document triage. It is not "
+    "legal, financial, or investment advice, and it does not replace review by qualified "
+    "legal, financial, or other professional advisers. Findings may be incomplete or "
+    "incorrect and must be independently verified against the source documents."
+)
+
+LIMITATIONS = """\
+- Analysis was performed by a local language model and may miss issues or misread context.
+- Scanned or image-only documents are not analysed (no OCR in this version).
+- Page references are approximate for findings that span page boundaries.
+- Only documents in supported formats (PDF, DOCX, TXT, MD) were reviewed.
+- Cross-document checks are based on extracted summaries, not full-text comparison.\
+"""
+
+
+def _finding_line(finding: Finding) -> str:
+    location = finding.source_file
+    if finding.source_page:
+        location += f", p. {finding.source_page}"
+    lines = [
+        f"- **{finding.title}** — severity: {finding.severity}, "
+        f"confidence: {finding.confidence:.0%}",
+        f"  - {finding.summary}",
+        f"  - Source: `{location}`",
+    ]
+    if finding.evidence.strip():
+        lines.append(f'  - Evidence: "{finding.evidence.strip()}"')
+    return "\n".join(lines)
+
+
+def render_markdown(
+    findings: list[Finding],
+    documents: list[str],
+    empty_documents: list[str] | None = None,
+) -> str:
+    high_priority = [f for f in findings if f.severity in ("high", "medium")]
+    counts = {s: sum(1 for f in findings if f.severity == s) for s in SEVERITY_ORDER}
+
+    parts = [
+        "# Vespera Due Diligence Review",
+        "",
+        f"*Generated on {date.today().isoformat()} · All analysis performed locally*",
+        "",
+        DISCLAIMER,
+        "",
+        "## Executive Summary",
+        "",
+        f"Vespera reviewed **{len(documents)} documents** and recorded "
+        f"**{len(findings)} findings**: {counts['high']} high, {counts['medium']} medium, "
+        f"{counts['low']} low severity, and {counts['info']} informational.",
+        "",
+    ]
+    if high_priority:
+        parts.append(
+            "Items flagged for priority attention: "
+            + "; ".join(f.title for f in high_priority[:6])
+            + "."
+        )
+        parts.append("")
+
+    parts += ["## High Priority Findings", ""]
+    if high_priority:
+        parts += [_finding_line(f) for f in high_priority]
+    else:
+        parts.append("No high or medium severity findings were recorded.")
+    parts.append("")
+
+    parts += ["## Findings by Category", ""]
+    for category in CATEGORY_ORDER:
+        in_category = [f for f in findings if f.category == category]
+        if not in_category:
+            continue
+        parts += [f"### {category[0].upper() + category[1:]} ({len(in_category)})", ""]
+        parts += [_finding_line(f) for f in in_category]
+        parts.append("")
+
+    parts += ["## Documents Reviewed", ""]
+    parts += [f"- `{name}`" for name in documents]
+    if empty_documents:
+        parts += ["", "Documents with no extractable text (possibly scanned images):"]
+        parts += [f"- `{name}`" for name in empty_documents]
+    parts += ["", "## Limitations", "", LIMITATIONS, ""]
+    return "\n".join(parts)
+
+
+def write_outputs(
+    findings: list[Finding],
+    documents: list[str],
+    output_dir: Path,
+    empty_documents: list[str] | None = None,
+) -> tuple[Path, Path]:
+    output_dir.mkdir(parents=True, exist_ok=True)
+    report_path = output_dir / "report.md"
+    findings_path = output_dir / "findings.json"
+    report_path.write_text(render_markdown(findings, documents, empty_documents), encoding="utf-8")
+    findings_path.write_text(
+        json.dumps([f.model_dump() for f in findings], indent=2), encoding="utf-8"
+    )
+    return report_path, findings_path

vespera-0.1.0.dist-info/METADATA

@@ -0,0 +1,158 @@
+Metadata-Version: 2.4
+Name: vespera
+Version: 0.1.0
+Summary: Local-first AI due diligence. Review a dataroom without sending confidential documents to a third-party AI provider.
+Project-URL: Homepage, https://github.com/VesperaSystems/vespera
+Project-URL: Repository, https://github.com/VesperaSystems/vespera
+Project-URL: Issues, https://github.com/VesperaSystems/vespera/issues
+Author: Vespera Systems
+License-Expression: Apache-2.0
+License-File: LICENSE
+Keywords: contracts,dataroom,due-diligence,llm,local-first,m&a,ollama
+Classifier: Development Status :: 3 - Alpha
+Classifier: Environment :: Console
+Classifier: Intended Audience :: Financial and Insurance Industry
+Classifier: Intended Audience :: Legal Industry
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Office/Business :: Financial
+Classifier: Topic :: Text Processing :: Linguistic
+Requires-Python: >=3.12
+Requires-Dist: httpx>=0.27
+Requires-Dist: pydantic>=2.7
+Requires-Dist: pymupdf>=1.24
+Requires-Dist: python-docx>=1.1
+Requires-Dist: rich>=13.7
+Requires-Dist: typer>=0.12
+Provides-Extra: dev
+Requires-Dist: build; extra == 'dev'
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Requires-Dist: twine; extra == 'dev'
+Description-Content-Type: text/markdown
+
+# Vespera
+
+**Local-first AI due diligence.**
+
+Review a dataroom without sending confidential documents to a third-party AI provider.
+
+```bash
+pip install vespera
+vespera review ./dataroom
+```
+
+Vespera scans the documents in a local folder — contracts, board minutes, NDAs — and produces a structured due diligence report with evidence-backed findings, each linked to its source document and page. All analysis runs on your machine via [Ollama](https://ollama.com). Document contents never leave your network.
+
+## What it does
+
+Vespera performs the *first-pass* review of a document collection for M&A, VC, and PE due diligence:
+
+- Recursively discovers PDF, DOCX, TXT, and Markdown documents
+- Extracts text and analyses it with a local LLM
+- Produces structured findings across categories including change-of-control clauses, termination rights, assignment restrictions, exclusivity, IP ownership, material liabilities, missing signatures, and cross-document inconsistencies
+- Writes a Markdown report and a machine-readable JSON evidence file
+
+```text
+Vespera
+
+Reviewing ./dataroom
+
+Documents found: 6
+Documents processed: 6
+
+Findings:
+- Change-of-control clauses: 1
+- Termination rights: 3
+- Missing signatures: 1
+- IP ownership / assignment: 2
+
+Report: vespera-output/report.md
+Evidence: vespera-output/findings.json
+
+All document analysis was performed locally.
+```
+
+Every finding carries its category, severity, a short verbatim evidence excerpt, a confidence score, and the source file and page.
+
+## Privacy model
+
+- **No cloud calls for analysis.** Inference runs on a local Ollama server (`localhost` by default).
+- **No telemetry, no accounts, no database.** Vespera reads your documents and writes two output files. That's it.
+- The only network activity you'll ever need is `ollama pull` to download a model once.
+
+## Quick start
+
+1. Install [Ollama](https://ollama.com) and pull a model:
+
+   ```bash
+   ollama pull qwen3:8b
+   ```
+
+2. Install Vespera (Python 3.12+):
+
+   ```bash
+   pip install vespera
+   ```
+
+3. Review a dataroom:
+
+   ```bash
+   vespera review ./dataroom
+   ```
+
+Try it on the included synthetic example:
+
+```bash
+git clone https://github.com/VesperaSystems/vespera
+cd vespera
+vespera review ./examples/sample-dataroom
+```
+
+### Commands
+
+```bash
+vespera review PATH [--model qwen3:8b] [--output vespera-output] [--host http://localhost:11434]
+vespera models      # show default + locally installed Ollama models
+vespera --version
+```
+
+## Supported document types
+
+| Format | Notes |
+| --- | --- |
+| PDF | Priority format; per-page source references |
+| DOCX | Paragraphs and tables; no page numbers |
+| TXT / MD | Plain text |
+
+Scanned image-only documents are not analysed in this version (no OCR).
+
+## Limitations
+
+Vespera is automated document triage. It is **not** legal, financial, or investment advice, and it does not replace review by qualified professionals. Local language models can miss issues and misread context; findings must be verified against the source documents. Vespera is designed to tell a human professional *where to look first* — not to make decisions.
+
+## Roadmap
+
+- OCR for scanned documents
+- More document formats (XLSX, EML, PPTX)
+- Additional local model providers (llama.cpp, MLX)
+- Configurable finding categories and custom review checklists
+- Multi-language document support
+
+## Development
+
+```bash
+git clone https://github.com/VesperaSystems/vespera
+cd vespera
+python -m venv .venv
+source .venv/bin/activate
+pip install -e ".[dev]"
+pytest
+```
+
+The LLM is behind a tiny provider interface (`vespera/llm/base.py`), so tests inject a fake provider and never require Ollama.
+
+## License
+
+[Apache-2.0](LICENSE)

vespera-0.1.0.dist-info/RECORD

@@ -0,0 +1,21 @@
+vespera/__init__.py,sha256=PFOxmfya9CVMO39ruE7LFTdXt7rK2w4pc21kbKjmgPk,68
+vespera/cli.py,sha256=dfGw_5St3zifWry-lXU1ku5y7R00KqfJJ7mtOj_tiDY,5162
+vespera/config.py,sha256=aIxkDvumITbWUs0Qdgb12lQGap19UW4-SDfrBf0T7ps,519
+vespera/documents/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+vespera/documents/docx_.py,sha256=-nqZxp8HS25wvLJXqQMzR3PBGxfwOWhJcgv-2h_rK0o,391
+vespera/documents/loader.py,sha256=_GH-iw-mOR8MWsoOthqpFTnGoFrNwf6dtP-ydPn8BpA,1425
+vespera/documents/pdf.py,sha256=u_NeBpSVBwvvGXzivDgQQ51UKrp5UAxA4ME1NSu_e88,356
+vespera/llm/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+vespera/llm/base.py,sha256=OrEjqbz9ibinT_bxHhzR9aKFrroING224f-2aDsNYI4,253
+vespera/llm/ollama.py,sha256=0MfhFNFvLX90yaGNc9qBzUVyBNLYjZdWPTFNHizGqkU,2338
+vespera/review/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+vespera/review/aggregator.py,sha256=bXbGCpmtLEA5jgNux86ghDY1a6aw4qTQcaTALuTZ7hg,1540
+vespera/review/analyzer.py,sha256=9s0rkS-eTOwnjUmAp3jMeGUTBOIf736HRKlequ2amug,5255
+vespera/review/models.py,sha256=7UjpsQm294eITJ8dkQKEPPkLGvLrJwaG7019rEelQ9E,2829
+vespera/review/prompts.py,sha256=NSzDdENwcrBk7R28CkRz-O0421wWCdmEpdh0ZOBoqcw,3563
+vespera/review/report.py,sha256=nDFS_0BMEg0yJA1x4Jm4BgIKvBid3S56xcuZf7T-SwY,4017
+vespera-0.1.0.dist-info/METADATA,sha256=mOYX6sBnGg1ut5ED7MbFyYlDUnAYTiMZ5JKfO763e6E,5070
+vespera-0.1.0.dist-info/WHEEL,sha256=lCkmxWfQsSc9CfIClYeavTdQeEX2toPqufh9gI35EQA,87
+vespera-0.1.0.dist-info/entry_points.txt,sha256=8yE1aNac0w7fVKyC6esee5bQkwlVJACz3vxK4X1-H04,44
+vespera-0.1.0.dist-info/licenses/LICENSE,sha256=z8d0m5b2O9McPEK1xHG_dWgUBT6EfBDz6wA0F7xSPTA,11358
+vespera-0.1.0.dist-info/RECORD,,

vespera-0.1.0.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: hatchling 1.31.0
+Root-Is-Purelib: true
+Tag: py3-none-any

vespera-0.1.0.dist-info/entry_points.txt

@@ -0,0 +1,2 @@
+[console_scripts]
+vespera = vespera.cli:app

vespera-0.1.0.dist-info/licenses/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.