Repository navigation
Help Assistant Internals 2 The Hybrid Search
Part 2 of 3. Part 1: from help pages to an index · Part 3: writing the answer, and learning from misses · Every design question, answered
A question arrives. In about a fifth of a second the assistant must decide which of 722 help sections answer it. It does that twice, two different ways, then merges the two:
- Word search - BM25, the classic search-engine scoring. Free, instant, no AI, exact about names.
- Meaning - the question's vector against every section's vector. Finds help that shares no words with the question.
- The merge - reciprocal rank fusion, words at half weight.
Neither alone is as good as both. This page goes through each with the real numbers from one question, then shows how the system decides how sure it is.
The running example, from prtest, as an administrator in English:
how do I create a ticketing department
function helpTokens(string $text): array
{
$words = preg_split('/[^\p{L}\p{N}]+/u', mb_strtolower($text), -1, PREG_SPLIT_NO_EMPTY) ?: [];
// ...drop words under 2 letters and common English ones (how, do, i, a, the, ...), then:
$out[] = helpStem($w);
}helpTokens('How do I create a ticketing department?') → ["creat", "ticket", "department"].
helpStem() is a cut-down Porter stemmer. It does not need to produce real words - only to make the question and the help agree, because both go through it:
| In | Out |
|---|---|
| departments | department |
| categories | category |
| statuses, status | status |
| adding, added, adds, add | add |
| settings, setting | set |
| create, created, creating | creat |
| route, routing | rout |
| class, analysis, business | unchanged |
TRAP: a stemmer's rules must chain. The first version returned after the first rule, so
settingsbecamesettingand stopped, never meetingset; andaddinglost itsingand then a doubledd, becomingad. Now: plurals, then-ing/-ed(undoubling a final consonant only when more than three letters remain), then a silente.
Non-English words are only lower-cased - that is the reason meaning, below, matters so much for other languages.
The first build scored whole sections, and the very first test question - this one - missed its answer. Tickets → Settings is about 10,000 characters of cards (the average section is 1,500), and BM25 rightly stops a long text winning just by being long, so its one Departments card was buried. Now each section is cut into passages of about 700 characters, made of whole blocks (helpPassages()), passages are scored, and a section scores as its best passage - which also becomes its preview.
On prtest that is 1,853 passages for 722 sections.
For each question word t in a passage d:
k₁ = 1.2 and b = 0.75 (the textbook values). N is the number of passages, nₜ how many contain the word, f how often it appears in this passage, |d| the passage's length in words against the average. Title words are counted three times (HELP_TITLE_WEIGHT), and a section from the module the asker is in gets ×1.15 (HELP_MODULE_BOOST) - a tie-breaker, never a filter.
The real numbers for our question:
| Word | Passages containing it (of 1,853) | idf |
|---|---|---|
ticket |
513 | 1.28 - in more than a quarter of the help; says little |
creat |
262 | 1.96 |
department |
99 | 2.93 - rare; narrows it right down |
That is the whole intuition of BM25: a word everyone uses barely counts, a rare word counts a lot, repetition helps with diminishing returns, and length is evened out.
The top of the word ranking:
| # | Section | Score |
|---|---|---|
| 1 | System help → Orphaned tickets → How tickets get orphaned | 9.11 |
| 2 | Tickets help → Settings | 8.76 |
| 3 | System help → Teams → Departments | 8.01 |
| 4 | Tickets help → Working with tickets | 7.89 |
| 5 | Workflows help → Actions | 7.78 |
Good, not perfect: orphaned tickets talks about deleting departments - all three words, wrong meaning. Words cannot tell "create" from "delete" near the same nouns.
helpRank() is pure - arrays in, arrays out - so all of this is tested without a database. Over 722 sections it takes about 80 ms in PHP.
An embedding model maps a text to a point in a space with hundreds of dimensions, placing texts that mean similar things near each other - whatever words they use, in whatever language. The question is embedded when it is asked (OpenAI's text-embedding-3-small, 512 dimensions, about 0.4 s), and compared with every candidate section's stored vector by cosine similarity - the angle between them:
function aiCosineSimilarity(array $a, array $b): float
{
$dot = 0.0; $na = 0.0; $nb = 0.0;
for ($i = 0, $len = min(count($a), count($b)); $i < $len; $i++) {
$dot += $a[$i] * $b[$i]; $na += $a[$i] * $a[$i]; $nb += $b[$i] * $b[$i];
}
return ($na == 0.0 || $nb == 0.0) ? 0.0 : $dot / (sqrt($na) * sqrt($nb));
}1 is the same direction, 0 unrelated. helpVectorRank() loads only the candidates' vectors (the analyst's language and English - not all 1,381), compares them all (about 16 ms), and ranks:
| # | Section | Similarity |
|---|---|---|
| 1 | Tickets help → Overview | 0.609 |
| 2 | System help → Teams → Departments | 0.606 |
| 3 | System help → Orphaned tickets → Reassigning | 0.577 |
| 4 | Tickets help → Settings | 0.560 |
| 5 | System help → Orphaned tickets → How tickets get orphaned | 0.551 |
Meaning puts Departments second, and pushes the orphaned section (the wrong meaning) down. Its own first choice, the Tickets overview, is close but generic.
Every question gets a nearest section - even nonsense. So similarities under 0.32 are discarded (HELP_VECTORS_FLOOR):
| Question | Best similarity |
|---|---|
| best pizza recipe | 0.25 |
| xyzzy plugh frobnicate | 0.25 |
| the weather in paris tomorrow | 0.21 |
| the lowest real question in the set | 0.38 |
Without the floor, "xyzzy" got five confident-looking help cards. With it, meaning has nothing to add and the reply is honestly "I couldn't find anything".
This is where meaning earns its keep. Word search cannot match "Hogyan hozzak létre jegykezelő részleget?" to English help - there are no shared words. Vectors place it near "Create departments like IT, HR and Finance" anyway, because the model learned both languages. (Part 3: the AI covers the rest by searching in English itself.)
Two rankings, on different scales (BM25 scores against similarities). Rather than try to compare the numbers, reciprocal rank fusion compares positions: each list adds weight ÷ (60 + position) to each section it ranks, and the sums are sorted.
foreach ($word as $i => $r) $fused[$r['section_key']] += HELP_MERGE_WORD_WEIGHT / (60 + $i + 1); // 0.5
foreach ($vec['ranked'] as $i => $r) $fused[$r['section_key']] += 1 / (60 + $i + 1);
arsort($fused);The 60 is the standard constant: it stops the very top of one list swamping everything. Words count half - the value the question set chose (below).
Our question, worked:
| Section | Words | Meaning | Fused | Merged |
|---|---|---|---|---|
| Teams → Departments | #3: 0.5/63 = 0.0079 | #2: 1/62 = 0.0161 | 0.0241 | 1st |
| Tickets → Settings | #2: 0.5/62 = 0.0081 | #4: 1/64 = 0.0156 | 0.0237 | 2nd |
| Orphaned → How tickets get orphaned | #1: 0.5/61 = 0.0082 | #5: 1/65 = 0.0154 | 0.0236 | 3rd |
| Tickets → Overview | (not in top 20) | #1: 1/61 = 0.0164 | 0.0164 | 6th |
The section that wins was first in neither list. It was strong in both - which is exactly the signal that two independent methods agree. The word ranking's false friend (orphaned tickets) drops to third; meaning's generic first choice (the overview) drops off the top five entirely. Both right answers are now first and second.
A section only meaning found still makes the list - with the start of its text as the preview, since there is no matched passage.
Each merge was run over the cached vectors of the question set (no API cost):
| Merge | Plain (47) | Exact (6) | Vague (6) | Other language (6) | Right section first |
|---|---|---|---|---|---|
| Equal weight | 46 | 6 | 5 | 3 | 43 / 65 |
| Words × 0.5 | 46 | 6 | 6 | 3 | 46 / 65 |
| Skip weak words | 45 | 6 | 5 | 3 | 43 / 65 |
| × 0.5 and skip weak | 45 | 6 | 6 | 3 | 46 / 65 |
The search returns the five best. Without a chat AI, the window replies with a short canned line and those sections as cards - and the line must never sound surer than the search is. helpAssistantReplyKind():
| Kind | Words only | With vectors | Example line |
|---|---|---|---|
none |
nothing matched | nothing matched | "I couldn't find anything for that - try different words?" |
weak |
top score < 9 | similarity < 0.45 and word score < 9 | "These might be close:" |
clear |
top ≥ 1.25 × the runner-up | meaning's first choice leads the next by ≥ 0.10, or first by words and by meaning at once | "This looks like exactly what you need:" |
module |
several, best from the asker's module | same | "From the help for the module you're in:" |
several |
otherwise | otherwise | "Here's what I found in the help:" |
Every threshold came from the question set's real scores, not a guess. With vectors on, clear winner came up 37 times: a right section first 32 times, in the top five the other 5 - never wrong. Our department question is several - correctly: its two right answers are close together, and so is a wrong one.
The lines live in the language files (help_assistant.phrase.<kind>.N), as many variants as wanted, picked at random - and none claims an AI read anything.
Matching by meaning needs the embedding model right now, for the question. Every way that can fail falls back to words alone, silently:
| Situation | Result |
|---|---|
| No embedding model set up, or switched off | words |
| The provider cannot be reached, a key refused | words (logged) |
| A new model chosen, its vectors not built yet | the old model keeps answering |
| A rebuild to a new model fails | nothing changes |
| Below the floor | words |
TRAP: embed the question with the model IN USE, not the model chosen. Once an admin saves a new model, the saved settings describe the new one - but the vectors answering were made by the old one. The first version embedded questions with the saved settings, so choosing a model silently dropped the assistant to words until the rebuild finished. The model in use now keeps its own settings (
help_vectors_active_cfg, its key encrypted), written at the moment a rebuild switches over;helpVectorsReady()embeds questions with those.
tests/help-search-questions.php holds 65 realistic questions, each with every section that would be a right answer, in four kinds: plain ("how do I create a ticketing department"), exact names ("Warbot", "MoSCoW"), vague - sharing no words with the answer ("my boss wants a weekly PDF of how we are doing"), and other languages (German, Spanish, Hungarian, Polish). Every step of the build was measured against the same set:
| Plain (47) | Exact (6) | Vague (6) | Other language (6) | |
|---|---|---|---|---|
| Words alone | 43 | 6 | 2 | 2 |
| Meaning alone | 46 | 6 | 4 | 4 |
| Merged | 46 | 6 | 6 | 3 |
| + a chat AI writing the answer (part 3) | 46 | 6 | 5 | 3 + 1 honest "not covered" |
A right section in the first five; for the AI row, the answer cites a right section.
TRAP: help that quotes the test questions flatters the measurement. The help assistant's own System help page uses "my boss wants a weekly PDF" as an example - so word search now ranks that help page first for the very question meant to test meaning. The merge still finds Report packs (meaning ranks it second), but a vague question's word score is no longer honest. When writing help for the assistant itself, use examples that are not in the question set - or add the question set's wording to the list of things help pages avoid.
FreeITSM — an open-source IT Service Management platform · github.com/edmozley/freeitsm · MIT licence
- Installation
- ⏰ Scheduled tasks (cron jobs)
-
Architecture
- ↳ 🧹 Root folder tidy (URL routing)
- ↳ 🔗 Database Integrity (FKs & constraints)
- ↳ 🛠️ Database Verification — Developer Guide
- ↳ 🧩 Service Layer (UI + API share code)
- ↳ 🪤 Code traps (the TRAP: comments)
- ↳ 📝 Commenting for the Engineering wiki
- ↳ 🛠️ Engineering wiki scanner - Developer Guide
- ↳ 📊 Service-Layer progress
- 🤝 Contributing (read before a pull request)
- 🧪 Developer tests
- 📄 PDF exports moved to PDFKit (#254)
- AI Providers
- Internationalisation (i18n)
- Timezones & Time Handling
- 📅 Date & Time Formats
- 💱 Currencies
- 🧭 The sidebar (optional module rail)
- 🧩 Preference templates
- 🔀 Display order (modules, Watchtower cards) — Developer Guide
- Theming & Dark Mode
- 🗂️ Recent — getting back to what you were doing
- ⌨️ Command palette (⌘K)
- 🔍 Searching inside tickets
- 📄 Attached documents
-
Mobile‑Friendly
- ↳ 🎫 Mobile: Tickets
- ↳ 💻 Mobile: Assets
- ↳ 📅 Mobile: Calendar
- ↳ 📚 Mobile: Knowledge
- ↳ 🚦 Mobile: Service Status
- ↳ 🗼 Mobile: Watchtower
- ↳ 🧩 Mobile: Problem Management
- ↳ 🔁 Mobile: Change Management
- ↳ 💿 Mobile: Software
- ↳ ✅ Mobile: Tasks
- ↳ 📝 Mobile: Forms
- ↳ 📄 Mobile: Contracts
- ↳ 📄 Mobile: Domains
- ↳ 📄 Mobile: People
- ↳ 🚀 Mobile: Projects
- ↳ 🗂️ Mobile: Files
- ↳ 🎓 Mobile: LMS
- ↳ 🗺️ Mobile: CMDB
- ↳ 🗺️ Mobile: Network Mapper
- ↳ 🧭 Mobile: Process Mapper
- ↳ ⚙️ Mobile: Workflow
- ↳ 🖥️ Mobile: System
- ↳ 📊 Mobile: Reporting
- ↳ 📖 Mobile: System Wiki
- ↳ 🙋 Mobile: Self-Service Portal
- ↳ 🧰 Mobile: Techniques & Tricks
-
Security
- Layer 1 — which modules you can enter
- ↳ 🧩 Module Access Control
- ↳ 🛠️ Module Access — Developer Guide
- ↳ 🔌 Switching modules off
- ↳ 🛠️ Module availability — Developer Guide
- ↳ 📐 Module availability — design (#217)
- Layer 2 — what you can administer
- ↳ 🎭 Roles & Permissions
- ↳ 🛠️ Roles — Developer Guide
- ↳ 🔤 Why capabilities are constants
- Layer 3 — the System module
- ↳ 🔑 Admin Access Control
- Hardening
- ↳ 📄 Security review response 2026-08
- ↳ 🛡️ Security hardening 2026-08
- ↳ 🛠️ Security hardening 2026-08 — Developer Guide
- ↳ 🛡️ Round three — plain English
- ↳ 🛠️ Round three — Developer Guide
- ↳ 🛡️ CSRF protection (S4) — Developer Guide
- Single Sign-On (SSO)
- 🗂️ LDAP & Active Directory
-
🟦 Microsoft Entra ID
- ↳ 📥 Importing people and groups
- ↳ 💻 Intune: assets to their primary user
- ↳ 🏢 Organisation chart
- ↳ 🛠️ Entra directory sync — Developer Guide
- ↳ 🛠️ Entra import - review (#235)
- ↳ 🛠️ Entra follow-ups - review (#237-#240)
- ↳ 🛠️ Intune primary user - review (#236)
- ↳ 🛠️ Intune after an Entra import - review (#334)
- ↳ 🛠️ Intune primary user history - review (#338)
- ↳ 🛠️ Intune asset types - review (#339)
- ↳ 🛠️ Intune scheduled syncs - review (#344)
- ↳ 🛠️ Microsoft 365 in Integrations - review (#352)
- ↳ 🛠️ Organisation chart - reviews (#242, #265)
- 📇 CardDAV contact sync
- Browser Extension
- API Reference
-
🔌 REST API — how it works
- ↳ 📋 Proposal: custom fields, inventory import and change evidence (#194)
- ↳ 🛠️ Custom fields, evidence + inventory - review (#195-#197)
- ↳ 🛠️ Long forms in a task's Work view - review (#198)
- ↳ 🛠️ Empty Procedures fold - review (#247)
- ↳ 🎫 REST API: Tickets
- ↳ 💻 REST API: Assets
- ↳ 🔴 REST API: Problems
- ↳ 🟠 REST API: Changes
- ↳ 📚 REST API: Knowledge
- ↳ ✅ REST API: Tasks
- ↳ 🗄️ REST API: CMDB
- ↳ 📜 REST API: Contracts
- ↳ 🗓️ REST API: Calendar
- ↳ 💿 REST API: Software
- ↳ 🌐 REST API: Domains
- ↳ 🚦 REST API: Service Status
- ↳ ☀️ REST API: Morning Checks
- ↳ 📝 REST API: Forms
- ↳ ⚙️ REST API: Workflow
- ↳ 🏷️ REST API: Cost centres
- ↳ 🏢 REST API: Companies
- ↳ 📒 REST API: Chart of accounts
- ↳ 📁 REST API: Projects
- ↳ 🛡️ API threat protection
- ↳ 🧑💼 REST API: Analysts
- ↳ 🗺️ REST API: Network Mapper
- ↳ 🧭 Using the API docs page
- ↳ 📐 OpenAPI specification
- ↳ ✅ OpenAPI: kept correct
- ↳ 🛠️ Maintaining the catalogue
- Watchtower
-
Tickets
- ↳ Reply subject and replying to one message
- ↳ 🔄 Live updates
- ↳ 🛠️ Watching a ticket - Developer Guide (1 of 2)
- ↳ ✉️ Ticket address suggestions - Developer Guide
- ↳ 🛠️ CSAT delay and reminder - Developer Guide
- ↳ 🛠️ Ticket checklist step types - Developer Guide (1 of 2)
- ↳ 🛠️ Right-click Attach checklist - Developer Guide
- ↳ 🛠️ Checklists in the REST API - Developer Guide
- ↳ 🔗 A ticket's related block and attachments - Developer Guide
- ↳ 🛠️ Inbox.js map — Developer Guide
- ↳ 🛠️ Floating windows & the file viewer — Developer Guide
- ↳ 🛠️ Tickets settings — Developer Guide
- ↳ 📐 Simpler ticket and task forms - design (#218)
- ↳ 🗂️ The ticket workspace
- ↳ 📋 Forms on tickets
- ↳ 📊 The ticket table
- ↳ 📋 Rota copy and paste — Developer Deep Dive
- ↳ ✅ Checklists & SOPs
- ↳ ☑️ Mandatory fields
- ↳ 🧾 Custom ticket fields
- ↳ 🏷️ Ticket categories
- ↳ 🧾 A type for every new ticket
- ↳ 👥 Assigning tickets to a team, and escalation
- ↳ 🏢 One board across every company
- ↳ 🔎 Many companies: searchable lists (#358)
- ↳ 📊 Ticket dashboard — Developer Guide
- ↳ Mailbox Authentication
- ↳ 📤 Email send log
- ↳ ✉️ Emailed reports
- ↳ 📎 Attachment size limits and large emails
-
↳ Basic IMAP mailboxes
- ↳ 📐 Basic IMAP without the PHP extension (#257)
- ↳ 🛠️ Basic IMAP client — Developer Guide
- ↳ 🧪 Basic IMAP client — Testing and verifying
- ↳ 🛠️ Basic IMAP - Sent copy - Developer Guide (#270)
- ↳ 🛠️ Verifying a mail folder — Developer Guide
- ↳ 🛠️ Mailbox providers — Developer Guide
- ↳ 🛠️ How Gmail mail collection works — Developer Guide
- ↳ Email rendering & images
- ↳ SLA Management
- ↳ WhatsApp channel
-
↳
✈️ Telegram channel - ↳ ⭐ CSAT company scope and filters — Developer Guide
- ↳ 👥 Microsoft Teams channel
- ↳ 🗨️ Mattermost channel
- ↳ 💬 Web chat channel
- ↳ 🟣 Slack channel
- ↳ 🔗 Linking tickets
- ↳ ⓘ Record previews
- ↳ 📝 Ticket notes: internal or shared
- ↳ 🗒️ Canned responses
- ↳ ✉️ Limiting replies to particular senders
- ↳ 📨 Telling the analyst a ticket is theirs
- ↳ ✍️ Email signatures
- ↳ 🌐 The public web address
- ↳ 🔢 Ticket numbering
- ↳ 🙋 Raising a ticket for someone else
- ↳ 🔀 Merging tickets
- ↳ 🔒 Confidential tickets
- ↳ 👥 Portal managers
- ↳ 👁 Who has seen a ticket
- ↳ 📜 Reading long tickets
- ↳ ⑂ Splitting tickets
- ↳ ✅ Selecting several tickets
- ↳ 🗂️ The folder pane
- ↳ 🔽 Just my tickets, or no closed ones
- ↳ 🛠️ Snoozing tickets — Developer Guide
- ↳ 👥 Collision detection
- ↳ ⏱️ Time tracking
- ↳ 📄 Case reports - a ticket and its work as one PDF
- ↳ 📅 Scheduled work in your own calendar
- Problem Management
- Tasks
- 🚀 Projects
-
Assets
- ↳ 🛠️ Asset list by type - review (#340)
- ↳ 🏢 Moving an asset between companies
- ↳ 📍 Shared asset locations
- ↳ 🧑💼 Assigning assets to analysts
- ↳ 📆 Warranty and lease alerts
- ↳ 📄 Leases
- ↳ 🔭 Saved table views
- ↳ 🖨️ Recording anything, and importing it
- ↳ 🧩 Asset type layouts
- ↳ 🕒 Inventory freshness
- ↳ 🔗 Linked configuration item
- ↳ 🔌 Connected assets
- ↳ 🔍 Asset review
- ↳ 🏷️ QR asset labels
- ↳ 📋 Who holds what, and handover documents
- ↳ 🖥️ The inventory agent (PowerShell)
- ↳ 🗄️ Proxmox VE servers
- ↳ ☁️ VMware Cloud Director servers
- ↳ 🪟 Hyper-V and Failover Clusters
- ↳ 🔗 Linking equipment to tickets
- ↳ ☑️ Follow-up tasks on a ticket
- Knowledge
- Change Management
- Calendar
- Morning Checks
- Reporting
- Software
-
Forms
- ↳ 🎨 The form designer — Developer Guide
- ↳ 📐 Layout & the grid — Developer Guide
- ↳ 🗂️ Collections — grouping submissions
- ↳ 📄 Submissions as PDFs
- ↳ ⚡ What happens next — a form's own actions
- ↳ 🛠️ Sections & conditional logic — Developer Guide
- ↳ 🛠️ Lookup fields — Developer Guide
- ↳ 🛡️ Catalogue request approvals
- People
- Starters & leavers
- Domains
- 🗂️ Files
- Contracts
- Service Status
- 🔔 Notifications
- 🚨 War Room
- Self-Service Portal
- LMS
- Process Mapper
-
CMDB
- ↳ 🧱 Designing your CMDB
- ↳ 🛠️ CMDB — Developer Guide
- ↳ 🛠️ Impact analysis — Developer Guide
- ↳ 🛠️ Data quality — Developer Guide
- ↳ 🛠️ Record references — Developer Guide
- ↳ 🛠️ CMDB assistant - Developer Guide
- ↳ 🔗 Linked assets
- ↳ 📐 Linking an asset to its CI - design (#201)
- ↳ 📐 Discovery and redundancy-aware impact - design (#199)
- ↳ 🛠️ Relationship wording - review (#297)
- Network Mapper
-
Workflows
- ↳ 🛠️ Workflow engine — Developer Guide
- ↳ 🛠️ Workflows for tasks — Developer Guide
- ↳ 🛠️ Workflow notes on a ticket — Developer Guide
- ↳ 🛠️ Workflow category routing — Developer Guide
- ↳ 📐 Workflows in order, child tickets - design (#304)
- ↳ ⏰ Time-Based Triggers
- ↳ 🪝 Webhooks
- ↳ 💬 Webhook Message Formats
- ↳ 🔒 HTTPS Certificates & CA Bundles
- ↳ 🛠️ SSL Verification — the Global Setting
- ↳ 🐞 Workflow & Webhook Pitfalls
- Issue trackers (Jira, Azure DevOps)
- 📡 Zabbix (monitoring)
- 💬 Microsoft Teams alerts
- 📞 3CX (phone system)
-
System
- ↳ 🛠️ System cards by topic - review (#347)
- ↳ 🟦 Microsoft 365 (Entra ID and Intune)
- ↳ 🛠️ Preferences — Developer Guide
- ↳ 🧹 Maintenance
- ↳ 💾 Storage usage
- ↳ ✉️ Emailed reports
- ↳ 🔎 Finding people in the Analysts list
- ↳ 🚪 When an analyst leaves
- ↳ 🛠️ System → Branding — Developer Guide
- ↳ 🩺 Page health (D019) — Developer Guide
- ↳ ⭐ Feature Bingo
- ↳ 📍 Locations
- ↳ 🏷️ Cost centres
- ↳ 📒 Chart of accounts
- ↳ 📷 Photo Album
- ↳ Demo Data
- ↳ 🛠️ Demo Data — Developer Guide
-
Overview
- ↳ 📊 Progress tracker
- ↳ Concepts & vocabulary
- ↳ Email routing & mailboxes
- ↳ Settings: global vs per-company
- ↳ Users & self-service
- ↳ Staff cross-company access
- ↳ One company supporting another
- ↳ 🏢 One board across every company
- ↳ Worked examples
- ↳ Pitfalls & gotchas
- ↳ Scope: what it's for
- ↳ 🛠️ Developer Guide (make a module multi-company)
- ↳ 🗄️ Case study: CMDB (a linked graph)
- ↳ 🧪 Test harness (prove it's isolated)
- What this is
-
🐞 Bugs resolved
- ↳ 💱 #386 Project stuck in an old currency
- ↳ 🖥️ #393 Header ran off the screen
- ↳ 🧩 #392 Project dialogs shared element ids
- ↳ 🖼️ #391 Knowledge images from your computer
- ↳ 🕗 #375 REST API wall-clock times
- ↳ 📅 #374 Calendar popup ran off the screen
- ↳ 🏢 #373 All companies: Changes, CMDB, Starters
- ↳ 🏢 #356 Assets People and All companies
- ↳ 🎓 #366 LMS demo data would not import
- ↳ 🔔 #362/#365 Team handover showed only a name
- ↳ 🗄️ #369 Inbox empty until Database Verification
- ↳ 🪆 #360 Links opened inside the ticket
- ↳ 🔗 #354 Workspace Links missed tasks
- ↳ 🔒 Task company checks in the API (5.5.0)
- ↳ 🔒 Asset holders through the API (5.4.0)
- ↳ 🔒 Files folder delete did not check inside (5.4.0)
- ↳ 🔒 The Intune tab did not check the company
- ↳ 📂 #332 Files viewer would not open
- ↳ 📎 #333 Note attachments downloaded
- ↳ 🪝 #326/#327 Form webhook Full record
- ↳ 📄 #325 Form PDF download failed
- ↳ 🔁 Small fixes in 5.1.0
- ↳ 🔁 CMDB relationships read backwards (5.1.0)
- ↳ 🔁 #301 CSAT delay was never applied
- ↳ 🔁 #281 Comments imported from Jira never arrived
- ↳ 🔁 #288 A change failed to open with Enforce approvals on
- ↳ 📚 Long non-Latin articles never got an embedding
- ↳ 🎫 #280 Email lookups scanned the whole table
- ↳ ✅ #266 Dragging a task told nobody
- ↳ 📚 #255 Article previews showed HTML codes
- ↳ 📚 #252 A link to a Knowledge article did not open
- ↳ 📄 #254 PDFs could not print Polish
- ↳ 📚 #250 Clicking a folder left the open article on screen
- ↳ 🎫 #260 The dashboard cog did nothing
- ↳ 🎫 #248 A department rule never matched on four routes
- ↳ 🔒 A directory account could be matched to someone else
- ↳ 🎫 #244 Reply went to the assigned analyst
- ↳ 🔒 Mailbox sign-in was not checked
- ↳ 🔒 Four weaknesses around signing in
- ↳ 🔒 A lookup answer could name any record
- ↳ 🎫 #227 A mailbox could not be deleted
- ↳ 🎫 #226 A picture on a note could not be closed
- ↳ 💻 #222 The Assets page never loaded in 4.3.0
- ↳ ⚙️ #214 A "contains" condition did not match capitals
- ↳ ⚙️ #216 A workflow email from a chosen mailbox always failed
- ↳ 🐳 The Docker image was missing PHP extensions
- ↳ 🌐 A failed DNS lookup read as a missing record
- ↳ 🔒 A link could change data
- ↳ 🔒 Access outlived its removal
- ↳ 🔒 Knowledge articles could run scripts
- ↳ 🔒 Knowledge writes ignored permissions
- ↳ 🔒 Company boundaries (Oct 2026)
- ↳ 🔒 Smaller gaps (Oct 2026)
- ↳ 🔒 Escaped text could leave its attribute
- ↳ 🔒 Task descriptions were not cleaned
- ↳ 🔒 Task documents ignored company scope
- ↳ 🔒 Repeating tasks ignored company scope
- ↳ 📬 #183 Mailbox checks overlapped
- ↳ 🌐 #188 Transfer lock shown off when unknown
- ↳ ✅ #184 Subtask names and Overdue
- ↳ 🧩 #186 Teams: the buttons did nothing
- ↳ 🔒 Text in a change could run script
- ↳ 🔒 AI settings changed without the AI permission
- ↳ 🚪 #178 Deleting an analyst failed
- ↳ 🎫 A team with no departments showed 0 tickets
- ↳ 🚀 Projects: adding a phase did nothing on Save
- ↳ 🔒 Leavers could still sign in
- ↳ 🖼️ Logo and courses broke on Apache with PHP-FPM
- ↳ 🔢 Chat tickets ignored your ticket numbering
- ↳ 📅 Dates shown as a dash, or in server time
- ↳ 🔒 Assets → Users showed people from other companies
- ↳ 🔒 Restricted analysts could read other modules' data
- ↳ 🖼️ Replies with a picture in the thread failed to send
- ↳ 📎 Reply attachments never reached the customer
- ↳ 🛠️ Outbound email attachments — Developer Guide
- ↳ 🔑 A global SSO provider was missing from the portal
- ↳ 🔀 Behind a proxy, the SSO redirect said http
- ↳ ✏️ The portal tagline moved when you saved it
- ↳ 🎨 The portal settings screen forgot what you saved
- ↳ 🛡️ The approvals inbox said "Error" and nothing else
- ↳ 📄 A table's answers were missing from the PDF
- ↳ ◉ A single-select column let you tick every option
- ↳ 📐 The portal ignored a form's field widths
- ↳ 📋 The tasks board stopped taking clicks
- ↳ 🗂️ #121 The index list is out of date after upgrading
- ↳ 📅 #133 The calendar subscription was empty
- ↳ 📋 #131 Tasks always reopened on the board
- ↳ 💥 #129 Every page returned HTTP 500 after upgrading
- ↳ 🐳 #127 A PHP warning above the System page
- ↳ 🕐 #126 Notes stamped with the server's clock
- ↳ 🌍 Storing every date in UTC
- ↳ 🚪 The portal was down for everyone signed in
- ↳ ⚙️ #120 Workflow notes could never be written
- ↳ ⚙️ #123 Three errors when running Database Verification
- ↳ 📝 #122 The description box was a stub in the corner
- ↳ 💣 Demo data deleted real accounts
- ↳ 🔐 #117 Sign-in redirected to the wrong address
- ↳ 🎨 #108 The priority dot was invisible
- ↳ ⏱️ #116 Time logged from the right-click menu
- ↳ 🔑 #114 API keys refused by our own guard
- ↳ 🗂️ #110 Assigning a task told nobody
- ↳ 🚪 #107 Signed out while still working
- ↳ 📎 #103 "Share with Requester" reached nobody
- ↳ 🔍 #102 Search found nothing for hyphens
- ↳ 🪟 #101 Source code editor opened behind
- ↳ ☑️ #88 Subtasks could not be ticked off
- ↳ 💻 #84 Asset deep link selected nothing
- ↳ 🎫 #79 A new ticket arrived with no status
- ↳ 📧 #79 A ticket from email did not say so
- ↳ 🔔 #78 Bell opened to nothing
- ↳ 📬 #77 Mail only collected from Inbox
- ↳ 🔐 #74 The default password could not be changed
- ↳ 🚦 #70 Renaming an impact level
- ↳ 📤 #67 App-only mailboxes could not send
- ↳ 📭 #45 Verify only ever worked for Microsoft
- ↳ 📭 #45 IMAP reported as not authenticated
- ↳ ✉️ An email template stopped escaping itself
- ↳ 🕐 The portal dashboard showed the wrong time
- ↳ 🔢 The folder said 99 and the list showed 96