Pipeline hub

Choose the right inference path, then dive into guides and notebooks

Pipelines

smol-doc-analyzer ships two complementary inference chains. Use this page as the map; each card links to design notes, CLI usage, and rendered notebooks.

flowchart TB
  Q{Need paper Fig. 1<br/>medical / salvage<br/>image-first flow?}
  Q -->|Yes| DICIE[DICIE · src/docie/]
  Q -->|No — need memos / ACORD intake| MEMO[Memo chain · src/pipeline/]
  DICIE --> DN[Notebook: DICIE walkthrough]
  MEMO --> MN[Notebook: Pipeline walkthrough]

When to pick DICIE vs the memo chain

DICIE — Document Image Classification & Information Extraction

DICIE

Design notes

Stages, taxonomies, review gates, FastAPI serve shape.

Notebook

DICIE walkthrough

Stage-by-stage demo with page images and stored outputs.

CLI

Memo chain — markdown → classify → extract → vision → summarize

Memo

Architecture

Chronological reaction contract and markdown conversion.

Notebook

Pipeline walkthrough

End-to-end memo-chain analysis with figures.

CLI

Supporting systems

System	Role	Links
Sample corpus	Queryable synthetic medical / salvage store	Guide · Walkthrough
Classical RF	TF-IDF + Random Forest baseline	Notebook
Discord (Chloride)	`/analyze` memo-chain front-end	Usage
Provenance	Synthetic-only disclosure	Data Provenance

--- title: "Pipeline hub" subtitle: "Choose the right inference path, then dive into guides and notebooks" --- ::: {.hero-kicker} Pipelines ::: ::: {.hero-lead} smol-doc-analyzer ships **two complementary inference chains**. Use this page as the map; each card links to design notes, CLI usage, and rendered notebooks. ::: ```{mermaid} %%| fig-cap: "When to pick DICIE vs the memo chain" flowchart TB Q{Need paper Fig. 1 medical / salvage image-first flow?} Q -->|Yes| DICIE[DICIE · src/docie/] Q -->|No — need memos / ACORD intake| MEMO[Memo chain · src/pipeline/] DICIE --> DN[Notebook: DICIE walkthrough] MEMO --> MN[Notebook: Pipeline walkthrough] ``` ## DICIE — Document Image Classification & Information Extraction ::::: {.feature-grid} ::: {.feature-card} DICIE ### [Design notes](../docie_pipeline.qmd) Stages, taxonomies, review gates, FastAPI serve shape. ::: ::: {.feature-card} Notebook ### [DICIE walkthrough](../notebooks/docie_pipeline_walkthrough.ipynb) Stage-by-stage demo with page images and stored outputs. ::: ::: {.feature-card} CLI ### [Usage · DICIE](../usage.md) `python -m src.docie` batch + response-only recipes. ::: ::::: ```{mermaid} %%| fig-cap: "DICIE stage order (paper Fig. 1)" flowchart LR A[Document Processing] --> B[Classification] B --> C[Information Extraction] C --> D[Aggregate / respond] ``` ## Memo chain — markdown → classify → extract → vision → summarize ::::: {.feature-grid} ::: {.feature-card} Memo ### [Architecture](../architecture.qmd) Chronological reaction contract and markdown conversion. ::: ::: {.feature-card} Notebook ### [Pipeline walkthrough](../notebooks/pipeline_walkthrough.ipynb) End-to-end memo-chain analysis with figures. ::: ::: {.feature-card} CLI ### [Usage · orchestrator](../usage.md) `python -m src.pipeline.orchestrator` and batch_runner. ::: ::::: ```{mermaid} %%| fig-cap: "Memo-chain chronological stages" flowchart LR A[to_markdown] --> B[classify] B --> C[extract] C --> D[vision_llm] D --> E[summarize] ``` ## Supporting systems | System | Role | Links | |--------|------|-------| | Sample corpus | Queryable synthetic medical / salvage store | [Guide](../sample_document_corpus.md) · [Walkthrough](../notebooks/sample_document_corpus_walkthrough.ipynb) | | Classical RF | TF-IDF + Random Forest baseline | [Notebook](../notebooks/random_forest_text_handwriting_classification.ipynb) | | Discord (Chloride) | `/analyze` memo-chain front-end | [Usage](../usage.md) | | Provenance | Synthetic-only disclosure | [Data Provenance](../data_provenance.md) | ::: {.see-also} ### See also [Quick Start](../quick-start.qmd) · [Commands](../reference/commands.qmd) · [Implementation plan](../plan/plan.md) :::