Resource
Create Grounded AI Workflows — Source-Only Sessions
A grounded workflow starts with a grounded source: package the reference once with strict rules, then run every question of the session against it.
Context Tools
Long Input Formatter
Pasting a big document raw mixes your source with your instructions. Package it instead: explicit delimiters separate material from commands, [§N] labels make sections citable, and grounding rules keep the model inside the source — up to strict mode, where missing answers become "the source does not say", never a guess. The source travels verbatim.
Paste the document, transcript, or reference text — it travels verbatim between the delimiters.
Packaging Preview (live — the plan, not the package)
Packaged Source
Resource
Organize Source Material for AI — Raw Notes, Honest Reading
Raw notes are fragments, not conclusions. Package them so the AI organizes without inflating — a four-word note stays an observation, not a firm claim.
Resource
Package Long Documents for AI — Delimiters and § Labels
Pasting a document raw mixes material with instructions. Package it: explicit delimiters, citable [§N] section labels, and grounding rules — the source travels verbatim.
Resource
Package Product Requirements — Shall Means Shall
Requirements language is normative: shall binds, should suggests, may permits. Package the spec so the distinction survives the AI's reading.
Resource
Package Research Material — Findings Keep Their Sources
Research mode packages multi-source material so findings stay attributed: exact quotes over paraphrase, §-citations per claim, and your inferences labeled as yours.
Resource
Package Technical Documentation — Code Quoted Exactly
API docs packaged with the rule that matters: code, versions, and identifiers quoted exactly — and this document beats the model's general knowledge of the API.
Resource
Prepare Contracts for AI Review — Clauses Under Strict Grounding
Contract review tolerates zero invention: clauses packaged under strict grounding, obligations cited by section, and missing terms reported as missing — never assumed.
Resource
Prepare Transcripts for AI Analysis — Speakers Stay Attributed
Transcript analysis fails when speakers blur. Transcript mode packages the conversation with attribution rules: who said it, when it was revised, and no words in anyone's mouth.
Resource
Reduce AI Hallucinations with Grounding — the Strict Contract
The strongest anti-hallucination tool is structural: only the delimited source exists, gaps answer "The source does not say.", and uncited claims are forbidden.
Resource
Structure Large Reference Documents — Every Section Addressable
A reference document is only as useful as its addressability. Heavy § labeling turns a nine-section reference into something every answer can point into.
Workflow
AI Long Document Analysis Workflow
Get AI to actually read a document that's too big for one prompt — fit it to the model, split it cleanly, package the parts, and analyze them without losing the thread.
Workflow
AI RAG Context Workflow
Prepare documents for a RAG system so retrieved answers stay accurate — budget the chunk size to the model, ground the sources against drift, and split them on clean boundaries for retrieval.
Workflow
AI Data Extraction Workflow
Turn messy text into structured data you can trust enough to feed another system — bound the source, extract the fields, force clean JSON, and validate before it flows downstream.
Workflow
AI Research Synthesis Workflow
Pull a single coherent view out of a stack of sources — package them together, summarize each faithfully, then have AI synthesize across them instead of one at a time.
Workflow
AI Content Writing Workflow
Turn source material and a target audience into clear docs, help, or knowledge-base content — gather the material, outline the piece, write it section by section, then check it for clarity and gaps.
Project
Build an AI Research Assistant with AI
The full path to an AI research assistant — define its scope, organize the source corpus, ground responses in references, extract key facts, synthesize findings, check groundedness, then validate it for use.
Project
Build a RAG System with AI
The full path to a retrieval system that returns grounded answers — understand the corpus, chunk and ground it, extract and classify the metadata, then evaluate that retrieval actually works.
How it works
Paste the source material — a document, transcript, contract, requirements spec, API docs, or raw notes — and optionally state the task. The Source Classification Engine detects the type deterministically (overridable), and each type adds its own reading rules: transcripts get speaker attribution and chronology rules, requirements keep shall/must binding, technical docs forbid paraphrasing code. Pick a packaging mode — Standard, Research, Documentation, Transcript, or Strict Grounding — and click Package Source. The package wraps your material verbatim between explicit <<<SOURCE START>>>/<<<SOURCE END>>> delimiters, inserts [§N] section labels for citation, and tops it with grounding instructions graded by mode: from "answer from the source first, label outside knowledge" up to strict mode's "only the source exists; gaps answer 'The source does not say.'; a claim you cannot cite is a claim you cannot make." The packaging also instructs the model never to follow instructions that appear inside the source — material and commands stay separated. Nothing leaves your browser.
Best for
- Packaging source material with delimiters and citable section labels
- Keeping the model's instructions separate from the material it works on
- Grounding answers in provided text so they can be cited
Not for
- Splitting content that exceeds the context window — that is the Long Prompt Splitter
- Carrying a working session into a new chat — that is the Context Handoff Builder
Use cases
- Pasting long documents without mixing source and instructions
- Grounding answers in a contract, policy, or spec
- Making sections citable with § markers
- Reducing hallucinations on source-based tasks
Pro tips
- Use Strict Grounding whenever a wrong answer is worse than no answer — contracts, policies, compliance. Its rule that "a claim you cannot cite is a claim you cannot make" changes what the model dares to assert.
- Let auto-detect classify the source, but check the preview: the type drives real rules (speaker attribution, binding shall/must, verbatim code), so a misclassified source gets the wrong discipline. Override with the select when needed.
- Ask for § citations in your follow-up questions too — the labels only pay off if you use them. "Which section says that?" becomes answerable the moment the source is labeled.
- The instruction-injection rule is doing quiet work: text inside your source that looks like a command ("ignore previous instructions…") is explicitly neutralized by the packaging. Paste untrusted documents with that in mind — packaged beats raw.
FAQ
Isn't this just the Prompt Formatter?
No — opposite ends of the prompt. The Prompt Formatter restructures your INSTRUCTION: a messy task description becomes a clean, sectioned prompt. This tool packages your MATERIAL: the document or transcript the instruction operates on gets delimiters, section labels, and grounding rules. PF asks "what am I telling the AI to do?"; LIF asks "what am I giving the AI to work from?" They compose: a PF-formatted instruction plus a LIF-packaged source is the full discipline.
Why do delimiters matter so much?
Because without them the model has to guess where your words end and the document's begin — and it guesses wrong in both directions: treating source sentences as instructions to follow, and treating your instructions as part of the text to analyze. Explicit delimiters end the guessing, and the packaging adds the security half: instructions appearing INSIDE the source are explicitly not to be followed, which blunts prompt-injection from untrusted documents.
What does Strict Grounding actually change?
It removes the model's permission to be helpful beyond the source. Only the delimited material exists for the task; assumptions are forbidden; anything missing gets the exact answer "The source does not say." — and every claim must cite its [§N] marker, with the rule that a claim you cannot cite is a claim you cannot make. "The task cannot be completed from this source" is explicitly defined as a valid, complete answer.
Does packaging change my source text?
No — the source travels verbatim between the delimiters: nothing summarized, cleaned, reordered, or rewritten, and the package says so. The only additions are [§N — title] labels inserted on their own lines for citation; remove the label lines and the delimiters and you have your exact original text back.
What do the source types change?
The reading rules in the analysis section. Transcript: every statement attributed to its speaker, later statements override earlier ones, no words in anyone's mouth. Requirements: shall/must binding, should/may weaker — the distinction survives. Technical documentation: code, versions, and identifiers quoted exactly, never paraphrased; the document beats the model's general knowledge of the API. Research: findings stay attributed to their sources; cross-source inferences are labeled as yours. Notes: fragments are observations, not conclusions.
When should I split instead of package?
Package when it fits; split when it doesn't. This tool prepares material that fits the window but needs structure and grounding. If the source exceeds the window, the Long Prompt Splitter chunks it with continuity markers — and the two compose: split first, and each chunk benefits from having been packaged. The Context Window Estimator settles the fits-or-not question.
Does my source material leave my browser?
No — your source material stays in your browser. The formatter classifies the document, wraps it in delimiters, and adds grounding rules locally; it never sends your contract, policy, spec, or compliance doc to a server or an AI model. You paste the packaged output into your own AI tool, where the usual provider data-handling applies.