Knowledge Hub

The Knowledge Hub is your team's place for questions, answers, and shared solutions. Think StackOverflow, but inside Spedy and integrated with your boards, tickets, and AI agents. Entries can be voted on, replies form a conversation thread, and an AI consolidation step turns useful threads into clean, versioned knowledge entries.

The Knowledge Hub is part of the Pro plan and is gated behind the AI_KNOWLEDGE_BASE organization feature.

How It Fits With the Wiki and AI Knowledge Store

Spedy now has three knowledge surfaces. They complement each other:

The Knowledge Hub and the AI Knowledge Store share the same underlying entries and the same hybrid search. A Knowledge Hub entry created by a team member is also visible to runner agents through knowledge_recall, and an entry created by an agent through MCP appears in the Knowledge Hub UI for humans to vote on, reply to, or consolidate.

Browsing Entries

Open Knowledge in the main sidebar. You see a paginated list of every non-deleted entry across your organization. Each entry shows:

• Title and a content snippet
• Category (Solution, Pattern, Convention, Insight, Preference, Error Fix)
• Tags
• Author and creation date
• Net vote score
• Reply count
• Linked board and ticket (if any)

Filters and Sorting

Use the filter sidebar to narrow the list:

• Category -- one of the six knowledge categories
• Tags -- comma-separated tag filter
• Board -- entries linked to a specific project
• Search -- hybrid search (PostgreSQL full-text + semantic vector)

Sort options include "Most Recent" and "Most Voted".

Entry Detail View

Click any entry to open its dedicated page at /knowledge/:id. The detail page is full-width and shareable by URL.

You see:

• The entry content rendered in Markdown
• Author, version number, last modified date
• Linked board and ticket
• Vote buttons (up / down) with the current net score
• A conversation thread with replies
• A "Version History" link

Replies

Replies form a flat conversation thread below the entry content. Both human users and AI agents can post replies.

Each reply:

• Renders Markdown via the same editor used for tickets and the wiki
• Has its own vote buttons and net score
• Shows author avatar, name, timestamp, and an "AI" badge when posted by an agent
• Appears in real time for everyone viewing the entry, via the existing Socket.io infrastructure

To post a reply, you need the ai-knowledge:create permission.

Mentioning AI Agents

You can @mention an AI agent team in a reply (using the same handles as in ticket comments). When the reply is submitted, Spedy creates a Runner job that processes the entry content plus the conversation thread as context. The agent's response is posted as a new reply, with the agent's team name as the author.

While the runner job is in progress, the entry shows a loading indicator so other viewers know an AI response is pending.

Voting

Each entry and each reply has a net vote score. Users can:

• Upvote -- click the up arrow; the score increases by 1
• Downvote -- click the down arrow; the score decreases by 1
• Toggle -- clicking the same direction again removes the vote
• Switch -- clicking the opposite direction removes the previous vote and registers a new one

Each user can cast at most one vote per entry or reply. Vote counts are visible to everyone with ai-knowledge:view permission.

Creating Entries

Click New Entry on the Knowledge page (requires ai-knowledge:create). The create form includes:

• Title -- short and descriptive
• Content -- Markdown editor with the same toolbar as wiki and tickets
• Category -- one of the six knowledge categories
• Tags -- free-text tags for filtering
• Linked board -- optional, scopes the entry to a project
• Linked ticket -- optional, attaches the entry to a specific ticket

Submit the form and the entry appears in the list immediately. AI agents can create entries through the existing MCP knowledge_store tool, and those entries appear in the same list with the agent as author.

Editing Entries

Users with ai-knowledge:edit permission can edit existing entries. Editing creates a new version automatically -- the previous content is preserved in version history.

AI Consolidation

When a knowledge entry has accumulated useful replies, you can consolidate the conversation into a new version of the entry.

How it works

1. Open the entry detail page
2. Click Consolidate (requires ai-knowledge:edit)
3. Optionally toggle Consolidate with AI to let an LLM merge the conversation into a new version
4. Confirm

Spedy reads the original entry plus the entire reply thread, generates a merged version of the entry content, and creates a new version. The version number increments to n+1.

Transactional safety

The LLM call runs before any database writes. If the LLM call fails, no partial state is written. All database writes for the consolidation happen inside a single transaction, with the version snapshot using upsert semantics so retries don't duplicate rows.

What happens to the old replies

The old replies remain visible in the conversation thread, with a visual divider marking "Consolidated into Version X" above them. New replies appear above the divider, so the timeline stays intact and nothing is lost.

Version History

Every change to an entry creates a version. Open the Version History panel from an entry's detail page to see:

• A list of all versions with timestamp, editor, and a short diff summary
• The full content of any previous version
• A Restore action that creates a new version from a previous one

Version history is read-only -- you can't delete versions, only roll back by restoring.

Search

The Knowledge Hub uses the same hybrid search as the AI Knowledge Store:

1. PostgreSQL full-text search -- exact lexical matches, BM25 ranking
2. Semantic vector search -- embeddings find conceptually similar entries even when keywords don't match

Results are merged and ranked by combined relevance. Search is rate-limited to 20 requests per minute per user.

Permissions

The Knowledge Hub is for internal roles only -- ADMIN and TEAM_MEMBER. Customers do not have access.

Permissions are managed through the standard Spedy permission groups.

API Endpoints

All endpoints live under /api/v1/knowledge and require an authenticated user with the appropriate ai-knowledge:* permission. The Knowledge Hub also requires the AI_KNOWLEDGE_BASE org feature to be enabled and an active Pro plan.

For full request/response schemas, see the OpenAPI spec.

Mobile Support

The Knowledge Hub is fully responsive. On phones and small tablets, the layout adapts automatically:

• The filter sidebar collapses into a horizontally scrollable chip row at the top of the page
• Entry cards stack vertically and take up the full screen width
• The detail view header wraps its metadata and action buttons to fit narrow screens

All features — searching, filtering, voting, reading replies — work the same way as on desktop. Only the layout changes.

Tips

• Vote actively -- the sort order depends on it; if good answers stay at zero, the list becomes useless
• Consolidate threads that come up twice -- if you've answered the same question more than once in a thread, that's the signal to merge it into the entry
• Link entries to projects and tickets -- linked context helps both humans browsing and runners using knowledge_recall
• Use categories deliberately -- "Solution" and "Pattern" are easy to search for; "Insight" is more freeform
• Mention agents in replies -- the existing mention system works inside knowledge replies, so you can pull an agent into a Q&A thread the same way you would in a ticket