Project Workspace¶
Start with the Project Workspace flow for the expected outcome, a practical sequence, and when to finish the discussion. This page describes the detailed controls and behavior.
Project Workspace is a feature-gated discovery area associated one-to-one with a Project. Its dedicated persistence model stays behind the normal Project UI.
There are two entry points:
- choose AI Workspace next to AI New to open an empty temporary workspace before a Project exists;
- choose Workspace on an existing Project to create or reopen that Project's workspace.
A temporary AI Workspace expires after 24 hours when abandoned and can be discarded explicitly. Discard workspace uses destructive styling and asks for confirmation before permanently deleting the workspace. Its expiry time is displayed in the browser's local timezone. When Genesis creates its Project, the same workspace becomes the persistent workspace for that Project. A Project cannot have more than one Project Workspace, and a workspace cannot create or attach to more than one Project.
Permissions¶
Authenticated users with normal Project read access can view a workspace. Analysts and Managers can create, edit, analyze, review, archive, delete, and promote workspace content. The AI Workspace and Project Workspace buttons are available to both roles when the feature is enabled. Assistant conversations are private to each user; Sources, Findings, and specification artifact drafts are shared within the tenant.
Source material and limits¶
You can add PDF, DOCX, Excel .xlsx, HTML, Markdown, and plain-text files, public webpage snapshots, and pasted text. Original uploaded binaries are not retained. The default upload limit is 20 MiB, webpage response limit is 5 MiB, and normalized text limit is 5,000,000 characters. Encrypted or text-empty PDFs are rejected. Archive expansion, PDF page count, parser memory, CPU time, wall time, and concurrent parser work are also bounded; a source that exceeds one of these safety limits is rejected without being imported. Analysts and Managers choose Add files to open a pane where they can drop one or more files or use Open files to select them with the operating-system file picker. The pane lists the selected files before they are added and, during upload, shows the current file and aggregate byte progress across the selection. Each accepted source also appears in Background jobs as AI Workspace source processing and progresses from queued to running to completed as its worker lifecycle advances. Select that job title to return to the AI Workspace that owns the source. The card header shows Ready after all enabled processing stages have completed successfully. The Ready Sources total uses the same rule.
Excel imports use cell values, including formula results saved in the workbook. Formula expressions are not retained in source text or evidence metadata, so new imports do not expose them in inspectors, evidence, search, or AI input. AIKOZO does not calculate formulas; a formula cell without a saved result is omitted. Recalculate and save the workbook in Excel before uploading it to include current results. Existing imports and historical answers are unchanged: upload the original workbook again to import values under this policy. Re-extracting knowledge from an old import does not remove its stored formulas. This policy also applies to Tenant evidence and other Sources using the shared Excel importer.
Source material is shown in the same card style as other application items and is not individually selectable. Most Sources are normalized as an immutable Source version → Page → Section → Block hierarchy. Excel Sources preserve the workbook structure as Source version → Worksheet → Row: each non-empty worksheet row remains one atomic processing and embedding block, including its column coordinates, and is never split into separate cell or partial-row chunks. During processing, one stage bar reports the active step and progress through structure, embeddings, knowledge extraction and knowledge projection. Once complete, only the header shows Ready. Cards show available Section and Block counts, or Worksheet and Row counts for Excel; hover over a count for its meaning. Unavailable counts are omitted. PDF cards also show their reliable physical-page count, while webpage cards show fetched-page and captured-link counts; document cards omit link counts. Completed knowledge extraction adds separate Concept and Assertion metrics. DOCX, HTML, Markdown, and text are flow-based formats: the processor preserves explicit DOCX page breaks and logical Page nodes, but it cannot reproduce layout-driven Word pagination without rendering the original file. Their cards therefore omit a potentially misleading page total, and their structure inspector does not repeat Page 1 on every item. Structure opens a spacious modal pane that lazily expands the hierarchy and captured links. Source actions are stacked in a right-side rail on wider screens and move below the content on narrow screens. Analysts and Managers can edit the title and content of pasted-text Sources or the title and URL of webpage Sources. A content or URL change stages a new immutable version and temporarily removes that Source from Assistant scope. A failed graph or vector projection can be retried from retained normalized content without uploading the original binary again.
When the Project Knowledge capability is enabled, a separate asynchronous extraction follows successful Source processing. The card header includes this enabled phase in its overall status. A Knowledge failure does not discard source text already processed. Extraction uses bounded Source windows and an operation-specific timeout, and the Source stage bar shows completed extraction units. A failed extraction stops; Retry knowledge repeats only the knowledge phase; it does not repeat structural graph or embedding work. Validated extraction units are checkpointed after a failure, so a compatible retry resumes those units and re-runs only missing or invalid units; a Source, prompt, model, schema, or unit-boundary change invalidates incompatible checkpoints. Failures appear as a small error glyph instead of expanding the card with technical output. Hover over the glyph for the full details or select it to copy those details to the clipboard. Knowledge opens the read-only Knowledge explorer for Concepts and Assertions scoped to the Source by default, with an all-Sources scope, evidence navigation, and an on-demand local graph. Each Expand another hop click adds one hop without a fixed depth limit; selecting another Concept starts at one hop again. The Concept list shows approximately eight rows before scrolling so the Local graph remains visible. Evidence opens in a pane above the Knowledge inspector, and Local graph can open in its own fullscreen pane above both. Assertions always link back to supporting Source elements; a failed extraction can be retried without reprocessing the Source.
When Knowledge Graph is enabled, focused Knowledge questions also retrieve the original passages supporting matching concepts and assertions. You do not need to request graph traversal explicitly. For example, a name found in an isolated table cell can be supplemented by the table that establishes the person’s role. Every passage still passes relevance assessment and remains within the selected source scope.
Webpage ingestion accepts only public HTTP(S) addresses, validates every redirect, sends no cookies or authentication headers, and rejects private, loopback, link-local, reserved, multicast, unspecified, and metadata destinations. Each connection is pinned to the public address that was validated for that request while HTTPS certificate validation continues to use the requested hostname. Semantic header, nav, and footer elements are excluded from extracted text and captured links so repeated page chrome does not enter source processing. The root page is followed to one click of unique same-origin HTTP(S) links only; child pages are never recursively traversed. robots.txt is honored and failure to retrieve it safely closes child traversal while still allowing the root page. The tenant setting Project Workspace crawl page limit controls the total number of pages including the root, from 1 to 100, with a default of 20. To change the limit to 5, an administrator opens Settings, opens the tenant, sets the value to 5 under AI Workspace source processing, and chooses Save. External, denied, capped, non-HTML, failed, and depth-limited links remain visible in the inspector.
The same Tenant Settings section controls the source chunk target (default 800 tokens), chunk overlap (default 80 tokens), and maximum knowledge extraction unit size (default 8,000 characters). Larger chunks and extraction units reduce embedding or extraction calls; smaller values keep retrieval and extraction more focused. Lower overlap reduces embedding volume. Saved values apply only when a Source is processed or reprocessed, so existing ready Sources are unchanged until then.
After section-level extraction, knowledge processing checks related sections within the same Source revision. Semantic similarity selects up to three other extraction units per unit; original passages must support every new or revised assertion. Definitions, explicit references, conditions and exceptions can connect sections or pages. Similarity alone never creates a relationship, and contradictory claims remain separate. This bounded search improves coverage but does not guarantee that every relationship is found. Progress shows related-section selection and cross-section checking before finalization. The evidence pane labels all contributing sections. Compatible retries reuse completed candidate embeddings and cross-section checks. Existing ready Sources change only when explicitly reprocessed.
Source-authored headings can be Knowledge evidence. For example, a person heading above their certificate entries supports a person Concept linked to the certificate Concept. Issue dates, expiry dates and status qualify that person’s certification relationship, rather than the shared certificate type. Assistant evidence previews retain the heading together with the cited passage. Existing Sources need explicit knowledge reprocessing after the corresponding application and prompt update; saved historical answers remain unchanged.
Grounding and citations¶
The AI Workspace Assistant is the primary discovery and structure workbench. Its persistent conversation is paired with three contextual views:
- Discover shows guided questions, exact workspace-state counts, and private pending structure changes;
- Decide contains the canonical shared Finding and decision inbox;
- Structure contains the Business and Technology Context drafts, generated Epic/Requirement hierarchy, validation, and promotion controls.
Full workspace opens the same workbench in a centred expanded layout without starting a separate conversation. Its title and Reset/Close actions form a centred horizontal group in the global header so the wider workbench begins immediately below it. The assistant conversation and inspector use an equal 50/50 split except on narrow mobile screens, where they stack. The AIKOZO logo, Help, and authentication controls remain visible. The conversation follows the standard AIKOZO AI Assistant interaction: Enter sends, Shift+Enter inserts a new line, the user message appears immediately, and an animated drafting message remains visible while AI is responding. Clicking a helpful question copies it into the message field so it can be reviewed or edited before sending. The primary Business Context question is What is the purpose of this project from consumer point of view? Before a structure draft exists, this appears as a helpful question alongside grounded prompts for drafting concise Business and Technology Contexts from the Sources. The list also includes Do we qualify to bid for this project call?, which asks the Assistant to compare mandatory requirements and prerequisites from Workspace Sources with people, certificates, skills, and competencies in published Organization evidence. The response cites gate and organization evidence separately, treats missing evidence as unknown rather than failure, and uses a conservative qualifies, does not qualify, conditional, or unknown conclusion. After generation, the Business Context prompt asks the Assistant to review and update the draft around that consumer-purpose question, while the Technology Context prompt checks the Workspace Sources and current structure for unsupported assumptions, missing constraints, and invented technology choices.
Assistant turns run in the background and resume their validated checkpoints after interruption. A conversation permits one active turn. Reopening the workspace resumes progress; failed turns expose Retry Assistant turn. Open results shows declared Sources, retrieval and processing coverage, uncertainty, and every saved item in pages of 50. Load more results continues to exhaustion; Ask about this item prepares a follow-up with a private result reference whose Source evidence is revalidated.
An explicit confirmation such as “Use that wording as the Project Business context” or “Set that as the Technical context” saves the selected prior Assistant wording to the linked Project. A Structure draft is not required. The Assistant removes conversation citation labels, preserves the approved wording in its original language, and maintains the Project's English/localized versions through the normal context save flow. Ambiguous requests need clarification; concurrent changes require a new save request. Progress shows Saving Project context and confirms the save when complete.
When an AI recognition or reconciliation response is invalid, the same job makes up to three attempts with validation feedback. Progress shows the current retry attempt. Valid independent responses are retained, and invalid batches do not add partial results. After exhaustion the turn fails visibly and Retry Assistant turn remains available. Source or configuration changes require a new turn.
Broad questions about the main user goals, gaps or unclear decisions in a selected specification read the whole selected specification, even without asking for “all” content. Targeted questions keep focused retrieval. Complete processing can take longer and uses bounded summaries and excerpts for the final answer. After a processing-contract change, an incompatible old turn requires a new message; its history and citations remain available, without re-uploading the Source.
Retrieval is intent-aware. Direct lookup searches literal identifiers in preserved text, headings and structure, keeping punctuation, Unicode and leading zeroes, and distinguishes exact, ambiguous and absent references. Complete collection requests enumerate every indexed content node in the declared Source versions with stable pages of 200 and verify terminal pagination and saved membership. Unqualified Workspace collections include every included Workspace Source. Published Organization evidence is added only when the detected intent visibly calls for organization, company, tenant, published, or shared evidence, or when the user explicitly selects or references it; Organization-wide questions then include the full permitted published collection. Explicitly named Sources constrain scope. Relevance can order processing but never shrink a complete collection, regardless of Source format or size.
Artifacts are recognized during inference, not classified at ingestion. Full text is processed in bounded batches (16,000-character target, subject to the configured model context budget); long content is split into addressable spans without truncation. Each processed span has a validated relevant, irrelevant or ambiguous outcome; failed processing remains visible and retryable. Recognized items stay private to the run. Similar items remain distinct unless their source boundaries establish the same item. The saved result count, not the chat summary, is authoritative. Exhaustive retrieval does not guarantee perfect semantic recognition.
Graph exploration supplements ambiguity resolution, relationships, cross-Source synthesis, rationale, consistency, impact and targeted gaps. The model chooses typed actions over permitted Concepts, Mentions and Assertions, then retrieves their underlying Source passages. The default budget is six rounds, configurable in Tenant Settings with batch and context sizes. Missing or incomplete semantic graphs permit Source-based reasoning with an explicit limitation. Exhaustive impact questions still scan every declared Source; graph budget exhaustion produces a partial reasoning answer, never an absence or completeness claim. Organization publication rules are unchanged.
Source material is untrusted data. Answers render supported Markdown and validated [S1] citation buttons open immutable Source text spans with version, checksum and document hierarchy. An answer that references a label without that validated mapping is rejected. Removed or revoked evidence cannot be reused. Existing historical messages remain readable when their evidence is still permitted. Explicit requested draft changes and Finding proposals run only after validated, supported processing and existing review/version checks.
The Assistant uses the active draft with recent conversation to interpret requests to create, save or update a reviewable proposal. Changed Structure items still require review and acceptance before promotion. A material ambiguity produces one focused question; an unavailable operation produces only its specific reason. Neither outcome starts evidence retrieval or changes the draft.
Follow-up questions use up to six recent complete user/Assistant turns, bounded to 24,000 normalized characters, to resolve pronouns, corrections, and omitted subjects before Source retrieval. Older messages remain visible in the conversation but are not copied into the inference window. The current request and current Workspace state always take precedence. Previous Assistant answers help resolve conversational intent only: they are not evidence, and their citations are reusable only when the same Source chunk is retrieved and allowed for the current response. Reset starts a new active conversation with empty inference history.
After each completed turn, the user message shows a muted, single-line Intent label of up to five words derived from the validated retrieval plan. It starts with the main operation, such as Retrieve all, Look up, Search, Summarize, Calculate, Compare, Update, or Assess, then narrows its scope. The Assistant interprets the complete request, including negation, compound instructions, and corrections, and supplies an operation, language, and short scope. The application presents consistent English or Slovak operation labels while preserving the scope phrase, human keys, and grammatical connecting words. Slovak input without diacritics is supported; accented names and English technical terms do not determine the label language. An explicit response-language request takes precedence, and a neutral follow-up retains the conversation language. Labels describe the requested work: calculating over every Source row still shows Calculate, even when retrieval is exhaustive. Hover over a shortened label to read its full text. The label is feedback; its validated intent scope also prevents retrieval from silently widening into Organization evidence. The label can retain an exact human identifier when useful, but does not repeat the prompt, expose a Source title, or claim retrieval succeeded. To correct it, send the correction as the normal next prompt. The copy glyph floats in the bottom-right corner of each message without adding an action row. It copies the original prompt from a user message and the response from an Assistant message. Hover over a completed message bubble to reveal its compact creation date and time in the browser's local timezone.
When an Assistant answer refers to a Finding, the conversation shows a short excerpt from the Finding statement instead of its internal database identifier. Hovering the excerpt shows the complete Finding statement. Internal identifiers remain available only to the structured lineage operations that need them.
Assistant proposals are saved immediately as Proposed Findings. They are never accepted automatically. Ordinary Findings require evidence. Gaps, open questions, and missing-information proposals may explicitly record that evidence is unavailable.
When a generated structure exists, retrieval planning receives a compact overview with its version and artifact counts. Source-only questions do not load the full draft. Questions about the draft and explicit edit requests review its complete Context/Epic/Requirement records and accepted Finding references in bounded pages, preserving fields, relationships and lineage. Progress shows Reviewing draft and reviewed-record counts; interrupted turns resume from their last validated page. The draft remains available in Structure. Proposed changes stay private until every page and the final answer pass validation. It can explain the structure, identify missing coverage or duplication, and make typed changes. Supported changes can add a missing Business or Technology Context, edit Context summaries, edit Epics or Requirements, move Requirements, add artifacts, merge duplicate Requirements, or split a Requirement. An explicit grounded request to change an identifiable Context, Epic, or Requirement updates the generated draft directly instead of returning only a descriptive answer or requiring a separate proposal approval. All valid changes from one Assistant response are validated and applied atomically with evidence lineage and optimistic version checks. The response displays Draft updated with an Open Structure action. The Assistant does not accept, reject, restore, or promote Structure items; the end user explicitly reviews and accepts affected drafts in Structure.
Artifact generation also checks completeness across reviewed inputs: every eligible reviewed requirement Finding must contribute lineage to at least one generated Requirement. A response that omits a reviewed spreadsheet-row requirement Finding is rejected and corrected before a draft is stored.
Draft specification artifacts¶
The validation panel omits repetitive per-item acceptance reminders. Artifact statuses and the Structure badge show what still needs acceptance; promotion remains blocked until all required items are accepted. Other validation errors remain visible.
The Draft specification artifacts step generates a structured, versioned candidate set after every Finding has been reviewed and every decision has been resolved. Generate structured draft launches a background job named Project structure generation, whose status changes from queued to running when its worker starts. Select that job title in Background jobs to return directly to the owning AI Workspace's Structure view. The set starts with exactly one Business Context and one Technology Context draft. Each is a concise, factual summary of no more than three sentences, grounded only in the reviewed Findings. The Business Context primarily answers what purpose the project serves from the consumer point of view; supported goals, actors, value, workflows, scope, and constraints clarify that purpose instead of replacing it with a feature list or technology description. When no technology decision or constraint is established, the Technology Context states that directly instead of inventing one. Epic candidates contain a key, name, objective, and scope. Requirement candidates contain a key, canonical English statement, and parent Epic. For a workspace associated with a Project, generation reads the localization selected in Project settings, requires every Context, Epic, and Requirement text in that locale, and displays that localized text in Structure while retaining canonical English internally. Promoted Requirements preserve both versions. Generated Requirement statements also use that Project's tenant format preset. A temporary workspace without a Project uses English and the tenant's default format preset. Reviewed Findings whose citations are unavailable, for example after their Source was deleted, remain available for audit but are ineligible for artifact generation and are omitted from the generated structure. At least one reviewed Finding with an available citation is required. Every generated candidate retains its own cited Source chunks and Finding lineage.
The Structure view presents the two Context drafts as full-width horizontal item rows, followed by one hierarchy in which each Epic is followed by its indented Requirements. Legacy drafts that predate Context generation show explicit Missing Business and Technology Context cards instead of an empty area. Its tab badge shows the number of proposed Contexts, Epics, and Requirements still awaiting acceptance. Every Context, Epic, and Requirement uses the same internal card layout as items in Decide: a full-width title and status row, content and evidence metadata below it, then a separated right-aligned action toolbar. Requirements remain visually nested through the indented, left-bordered list below their Epic. Analysts and Managers can accept, edit, reject, or restore Contexts and other candidates, move Requirements directly, or use Ask AI on a Context, Epic, or Requirement. The editor opens text in the Project locale; saving a localized edit updates that version and its canonical English translation atomically, so neither version can remain stale or be promoted inconsistently. A missing Context card asks the Assistant to add the missing grounded Context directly to the structure. Validation blocks promotion until both Context types are present, keys are unique, required fields are present, every active Requirement belongs to an active Epic, and every Context, Epic, and Requirement is accepted and evidenced. The view reports evidence coverage for the active artifact set.
For a temporary workspace, Continue in Genesis creates or reopens an idempotent Genesis session with the reviewed structure and lineage. After Genesis creates the Project and its initial Specification, the associated workspace can promote the accepted candidates into that Specification. For an existing Project workspace, the compact promotion card keeps the current layout when an existing Specification is selected. Choosing Create new Specification expands the card into separate full-width rows for the target selector, the new Specification key and name, and the promotion action; the action becomes available after both new values are entered.
Direct promotion runs as an idempotent background job shown as Promoting artifacts. Its status refreshes from queued through running to completion while the Background jobs pane is open. It writes the accepted Business and Technology Contexts at Project scope and creates draft Epics and Requirements through FastAPI, records candidate-to-artifact mappings, and reports completed and failed items. Retrying a partial failure skips already-created artifacts and retries only missing or failed items. A Genesis handoff carries the same accepted Context drafts into the new Project as Project-scoped contexts and creates its initial Specification.
Bilingual Specification artifacts¶
Contexts, Epic names/objectives/scope, and Requirement statements retain canonical English and the Project language throughout generation, Assistant edits, review, and promotion. Chat replies and Findings instead follow the language used to initiate the chat.
In the artifact editor, Editing language selects the authoritative version. Save queues Artifact language synchronization; both versions are saved together after translation succeeds. A failed translation leaves the stored artifact unchanged. If another edit changes the draft while synchronization runs, saving is rejected rather than overwriting newer work. English and localized fields have separate length limits, so a longer translation remains editable.
For older or conflicting drafts, select the authoritative language and enable Synchronize all text from the selected language. This explicitly replaces the other version with a fresh translation. No existing data is silently repaired. Validation identifies missing, mismatched, stale, or unverified translations and blocks acceptance/promotion until they are resolved. Changing the Project language requires generating a new structure.
Promotion preserves both languages in Project Business/Technology Contexts, Epic names/descriptions, and Requirement original/final text. Project Context and Epic editors also allow selecting the input language and synchronize its counterpart. Canonical English remains the input to internal reasoning and search. The translation prompt and model are configurable as Specification artifact translation in Settings and Tenant Settings.
Review and decisions¶
Analysts and Managers use the Decide view to accept, edit and accept, reject, or restore Findings. Its tab badge shows the number of Findings still requiring review or a decision. Finding cards use the full card width for the statement, keep type and evidence metadata together, and place actions in a compact toolbar below the content. Only proposed Findings and Findings needing a decision remain visible in the main review list. Accepted, edited, and rejected Findings are kept in the closed Reviewed Findings disclosure. Edit and accept opens the standard Finding modal; Escape cancels, Enter outside the multiline statement submits, and Enter inside the statement inserts a new line. Edited acceptance preserves the original proposal. Rejected Findings remain auditable. Open questions are Findings: they can be marked Needs decision, and Record decision opens the standard modal pane with a required decision note. In that pane, Escape cancels, Enter outside the multiline note submits, and Enter inside the note inserts a new line. Recording a decision preserves both the question and the decision note. Open questions are not duplicated as specification artifacts. Repeated assistant proposals with the same normalized Finding statement are not added again.
Auto approve asks for confirmation, then runs a background AIKOZO review over the exact current set of open Findings. AIKOZO uses the supplied evidence first and recognized best practices or recommendations for genuine gaps, then accepts, clarifies and accepts, rejects, or records a decision for every Finding. An open question or any Finding marked Needs decision can only receive a recorded decision or be rejected; the structured AI response schema prevents acceptance actions for those Findings. The complete structured decision set is validated before changes are applied. If the open Finding set changes while the job runs, the job stops instead of overwriting newer work. Each automated result keeps its AI action, rationale, and any decision note visible in the reviewed Finding for audit.
Deleting a Source removes its chunks and retrieval vectors, but historical Finding evidence remains as an unavailable snapshot instead of becoming a dangling link.
Linear workflow and Genesis¶
The assistant workbench keeps one continuous interaction across discovery, decisions, structure generation, structure changes, validation, and promotion. The progress path is Add Sources, Discover, Decide, Structure, and Promote. The slim Recommended next action strip identifies the current action and repeats its step number. The highlighted step and an aligned amber focus in the strip background provide the visual relationship without an arrow, connecting line, or shadow. Choosing Promote activates the Structure view and scrolls directly to Continue in Genesis for a temporary workspace, or to Promote accepted artifacts for an existing Project. Project candidates and standalone Project Drafts are not separate workspace steps.
When all Sources have been removed, Start here returns to the initial Add Sources state and resets its Findings, review, and draft counters. Retained Findings and drafts remain available as audit history in their workspace views, but they do not make the source-free workspace appear to have active discovery progress.
This Start here pattern is shared by the Specification, Implementation, and Feature workspaces. Each workspace keeps its own stages, KPIs, and actions while the placement and visual cues remain familiar. The content area associated with the recommended stage is also marked by a thin amber line on its left edge.
In a temporary AI Workspace, Continue in Genesis opens Project Genesis with the reviewed Epic/Requirement structure and lineage. A workspace for an existing Project promotes the same reviewed structure into an existing or newly defined Specification.
Non-goals¶
Project Workspace does not provide OCR, recursive or cross-origin crawling, object storage, external connectors, MCP resources, internet research by the assistant, ontology or terminology governance, inferred persisted facts, or implementation and delivery artifacts. It does not replace manual Project creation or the generic contextual assistant.
Personal Assistant prompts are available through the small ▴ glyph above the message composer or Arrow Up while editing a message. See AI Assistant prompts for reuse, favourites, retention and Profile management.
Prepared analysis and Assistant limits¶
Expand AI reading: 259 passages · 21 uncertain on a Source card to learn what the saved AI interpretations mean. Uncertain includes passages that could not be confidently interpreted or classified; it does not count missing Requirements or Findings awaiting review. Structure shows source previews, while Knowledge shows extracted Concepts and Assertions; neither lists the exact flagged passages or lets you mark them resolved. Ask about this Source prepares a source-scoped question for you to edit and send; ask something specific and check its citations. Questions do not clear the count. Counts appear only when available; older Sources may show AI reading details unavailable. Complete Source audits reuse compatible saved interpretations and examine original passages where needed.
“Am I missing anything in the project structure?” assesses one complete outline of current saved Business and Technology contexts, all Finding decisions, the selected artifact draft, current published Specifications/Epics/Requirements, and promotion links. It runs deterministic structure checks before asking the AI for an overall assessment. Normal readiness uses a planning call and an answer call, rather than an AI call for each small group of records. Changed project contents stop publication so an answer cannot silently describe an outdated state.
Progress calls this Assessing Project readiness. Open results shows how many Project records were included, saved structure checks, and whether the AI assessment finished. This is a current-structure assessment, not proof that every source requirement is represented. An explicit request to check every source requirement remains a complete audit.
The outline retains canonical artifact text, every Finding statement/status/decision, and source references. Full evidence excerpts and detailed locations remain available for inspection, outside this assessment. Compatible saved Source summaries provide background without re-extraction. A large outline that cannot fit the configured model context stops explicitly; records are never silently sampled. Older partial readiness turns require a new message after this release; saved history and evidence remain readable.
Focused questions share one query embedding across selected Sources and reuse the analyzed passages reached by search. Uncertain or unclassified passages still require original text. New Finding evidence records exact validated source spans; historical chunk-only evidence remains unverified rather than acquiring invented lineage. Saved Source summaries are produced during normal knowledge extraction; asking a question does not reprocess a Source.
Use Refresh analysis on a processed Source to apply the current analysis to its saved content, including generating summaries for older Sources. It preserves the Source, its stored structure and source embeddings; deleting or uploading the file again is unnecessary. The action is available to workspace editors when knowledge extraction is enabled and normalized content is ready. Progress appears on the Source card and in Background jobs; another refresh is unavailable while analysis is running. Refresh may use paid AI calls. Compatible checkpoints from interrupted runs are reused when available; completed runs do not retain those checkpoints. Refresh does not guarantee a lower uncertain-passage count. It also does not reimport the original file: use replacement when the original content or Excel import behavior needs to change.
The Assistant reports reused passages, original-passage work and draft/Project review progress. Unknown review totals are shown as pending. Invalid draft or answer output receives at most three attempts on the same page, retaining previously validated work. The default allowance per turn is 100 AI calls, 500,000 accounted tokens and 10 minutes, configurable under Tenant > Project Workspace. Accounted tokens include reservations for in-flight calls and calls with unknown usage; reported tokens are shown separately. This is not a currency spending limit.
Intermediate work reserves capacity for a final answer. A read-only turn reaching that reserve can return an explicitly partial answer with completed evidence and remaining gaps; it cannot propose changes or claim a complete audit. When an allowance is exhausted, Open results still provides completed work and any saved intermediate review summary. A validated answer can still be saved after the model deadline without another AI call. Retry does not reset the allowance and is unavailable for an exhausted turn. A new turn has its own allowance. Source, prompt or processing changes can require a new turn; compatible failed stages retain checkpoints for retry. Older processing contracts remain readable but cannot resume after this release.
If the time allowance expires during a model request, the turn reports Assistant work limit reached. A temporary provider connection error or timeout before that deadline follows the existing three-attempt retry limit on the current step; it does not restart completed review pages or reset the turn allowance.
The Assistant progress footer shows a compact status with result controls. Hover over or focus the information icon beside Open results to see coverage, AI calls, reported tokens and accounted tokens. Errors and retry actions remain visible.
The workspace loads only the latest full specification artifact draft when opened and while waiting for a newly generated draft. Older drafts remain stored; history listings return compact summaries, with full content retrieved separately for a selected draft.
Discover, Decide and Structure build their content when first opened. Returning to an unchanged tab preserves its current content; changes affecting a hidden tab appear when you next open it. Tab counts stay current even before that tab is opened, and the shared Assistant conversation remains visible across tabs.