Library

The Library is your organization's shared store of reusable documents — certificates, references, CVs, methodology templates, capability statements — that the AI can pull into any procurement. Upload a document once and the platform classifies it, indexes it, and makes it available to every future bid your team works on.

What the Library is for

Use the Library for documents that belong to your company and stay relevant across many tenders:

• ISO 9001 / 27001 and other certifications
• Reference letters and customer testimonials
• Specialist CVs and qualifications
• Past experience tables, capability statements
• Standard methodology and process descriptions
• Internal policies and procedures used as evidence
• Past RFPs and past bids worth reusing as templates

The AI reaches into the Library from four places: when it generates a preparation checklist, when it drafts a bid plan during the setup interview, when it composes a bid, and when it composes an RFP. When your library has the right material, the agent can attach existing documents to checklist items, pre-fill compose-plan attachments, and reuse content in drafts instead of asking you for each piece from scratch.

Library vs. Upload RFP

These are two different stores — don't mix them up:

If you upload a buyer's RFP into the Library by mistake, it will appear in every future procurement's search — remove it.

Adding documents

There are two ways to add documents to the Library.

From a bid or RFP document

Open any document inside a procurement — your own proposal docs, attached evidence, even past RFP material. In the document row's three-dot menu choose Save to Library. The platform creates a Library entry that points at the same file and automatically classifies it. The original document stays in the procurement; the Library entry links to that document rather than keeping an independent file copy. Replacing the source file immediately excludes its old search content and schedules indexing of the new content.

Procurement-sourced entries are auto-categorized — the RFP file becomes a Past RFP entry; everything else from a bid becomes a Past Bid entry.

This is the easiest way to grow your Library — every time you put together a bid, save the reusable parts so the next tender has them ready.

Directly in the Library

Go to Library in the sidebar and click Upload. You can select multiple files at once. Each file becomes a separate Library document and is classified by AI in the background.

When uploading, set:

• Name — a short, recognizable label
• Description — optional context (where it came from, when it was issued)
• Category — Reference, Past RFP, Template, Policy, Certificate, or Other

After upload, the AI fills in document type, standard / reference, issuing entity, issued to, valid until, a one-or-two-sentence summary, keywords, and a confidence score automatically.

Each file is capped at 50 MB.

Tabs: Standalone, From procurements, All

The Library page has three tabs that separate documents by where they came from. Each tab shows a count next to the label so you can see at a glance how much material you have in each bucket:

• Standalone — documents uploaded directly to the Library via the Upload button. These are the reusable assets you maintain deliberately.
• From procurements — documents that were saved into the Library from a bid or RFP using Save to Library. They still point at the original document inside its procurement.
• All — every Library entry, regardless of source.

Standalone is the default tab because that's the curated set most teams care about. Use From procurements to review what your colleagues have flagged as reusable from past tenders.

Supported file formats

Content search and AI classification are separate processes. A classification badge saying Format not supported does not mean the content cannot be indexed.

Scans and converted files may be only partially searchable. Check the original whenever layout, signatures or complete content matter.

Classification status badges

Every Library document carries a status badge. The status reflects what the AI has been able to do with the file:

Status updates live — the badge changes in real time as background classification finishes, so you don't need to refresh.

Content-search status

A separate Content search status appears alongside classification:

Choose Retry indexing for failed or partially searchable content. Retrying partial content keeps already published passages for the current source available while its replacement is prepared. Statuses update automatically.

Only documents added to the Library are indexed, not every procurement file. Content indexing itself does not deduct customer credits.

Searching your Library

Use the Search documents box at the top of the Library page to filter the visible list by name, description, or tags. The search is local to the page and updates as you type — useful when your Library grows past a screenful.

Agent search combines passages, exact text matches and metadata. Results can identify a page, paragraph or spreadsheet row; agents can expand a passage with nearby text. If content is not available, a metadata-only match may still be returned, explicitly without content evidence. Partial extraction and OCR errors require checking the original.

How the AI uses the Library

The Library is wired into four flows.

Preparation checklist generation

When you generate a preparation checklist for a procurement, the agent first calls library_search to discover documents that already cover RFP requirements, and uses them to inform what items it adds to the checklist. After generation, a separate matching pass runs against every attachable item — strong matches (confidence at or above 0.6) are auto-attached as suggestions. You'll see entries like "Suggested: ISO 9001 certificate.pdf — from your library" next to attachable items.

This means the more you invest in your Library, the less manual work each new tender takes.

Setup interview (bid plan)

During the setup interview the agent drafts each planned artifact in the compose plan, then matches your Library against the compliance coverage of that artifact. Suggested attachments appear next to each artifact for you to review and confirm before composition starts. Only confirmed attachments are pulled into the next phase.

Bid composition

When the bid composer starts, it first auto-pulls every confirmed Library attachment from the plan into its sandbox at /workspace/library/. From there the agent can also call library_search to find more material on demand and library_pull to bring extra docs into the same workspace. It can quote, summarize, or reference Library documents inside the generated proposal.

RFP composition

The RFP composer has the same library_search / library_pull pair, which it uses to find past RFPs, standard clauses, and templates from your organization's library when drafting a new tender.

The buyer still manually chooses the initial Library source for RFP composition. A retrieved reference does not confirm applicability or quote correctness.

Note: the bid-vs-RFP analysis agent does not use the Library. Analysis works only against the RFP and the proposal uploaded for that specific bid.

What belongs in the Library

Good candidates — reuse value high, content stable:

• Customer references and testimonials — letters of recommendation, named-client lists, case-study summaries
• Professional certifications — ISO 9001, ISO 27001, CBAP, vendor-specific certifications
• Specialist CVs — when the underlying experience is reusable across tenders
• Past experience tables — completed-projects matrices with dates, values, scope
• Standard methodology documents — how your team approaches delivery, support, security, training
• Capability statements and company overviews
• Past RFPs and past bids — when you'd reuse them as a starting point for future drafting

Bad candidates — single-tender, will go stale, or buyer-side:

• The current tender's RFP, technical specifications, or contract draft (those go under Upload RFP inside the procurement)
• Filled-in price/financial offer forms specific to one bid
• Tender-specific responses ("our approach for procurement X") that don't generalize
• Internal drafts that aren't approved for reuse

Library documents in the tender eligibility check

The tender eligibility check uses your classified Library documents as evidence, for example certificates, reference letters, CVs and financial statements. A document whose Valid until date has passed no longer counts, so keep certificates current. A requirement that could only be backed by an expired document is marked Not verifiable.

Tips

• Save aggressively from past bids. Every bid you finish, take a minute to save reusable evidence to the Library. Future you will thank past you.
• Upload EDOC archives directly. No more extracting inner files by hand — the platform unpacks the bundle and classifies each inner document for you.
• Name documents for retrieval, not for storage. "ISO 27001 certificate (valid until 2027)" beats "cert_v3_final.pdf".
• Set expiry-aware names. Putting the validity year in the name makes it obvious when something needs renewal — and the AI's extracted Valid until date will be flagged red on the show page once it has passed.
• Check the content-search status. Excel, CSV and legacy Word content may be searchable even when classification does not support the format. For partial results, check the original and retry indexing.
• Use categories. Sorting by Certificate, Reference, Template, Policy makes the Library navigable as it grows.

Archiving and source changes

Archiving immediately excludes a document from search; its index can remain for history and restoration. Replacing a source immediately excludes old passages, and a background job cleans up obsolete indexes. Hard-deleting a Library entry removes its passages and indexes. See security and privacy for stored content and AI processing.