Getting started

bitdispatcher is a computer-aided dispatch (CAD) system for 911 centres: one program with its own web interface. It creates and tracks incidents, recommends units by drive time over the road network, shows units, calls and incidents on a dispatch map with aerial photos, gives field crews a mobile terminal, and records every action in a tamper-evident ledger that reports are built from.

The first hour

1. Sign in as the administrator with the password printed on first start. The position is the console you are at.
2. Record the map: Admin › Map, choose the centre and radius, press Record. About two minutes for 100 miles. Open Map (chapter 46)
3. Stations and units: where units are based and what they are. Open Stations (chapter 44) Open Units (chapter 43)
4. Call types and run cards: what each kind of incident needs. The defaults are a sensible start. Open Call types (chapter 45)
5. Users: dispatchers, supervisors and field crews. Open Users (chapter 51)
6. Call handling: connect BitVoice, or point any call-handling system's ALI spill and EIDO at bitdispatcher. Open Integrations (chapter 49)

Demo mode

Started with -demo, bitdispatcher seeds a fictional agency on the Roanoke, Virginia map (stations at real station positions, made-up units). With bvsim, a simulated BitVoice, calls ring and radios report positions; everything simulated is marked SIMULATED.

The dispatcher screen

Everything a dispatcher needs is on one screen, and every part of it shows something useful.

Select an incident by clicking it, with [ and ], or from the palette (Ctrl K). Esc clears the selection.

Screens on several monitors

A position can spread its work over as many monitors as it has. Each monitor shows one screen: a window of the same live picture, arranged for one job. All of a position's windows share one sign-in and one selection: choose an incident on the queue screen and it opens on the incident screen and the map.

Open Screens with the monitor button on the status bar, from the command palette, or with Ctrl+Shift+S.

• Add a screen, choose its disciplines (all, fire, EMS, police, other) and Open it. Open all opens every one.
• Detect monitors (Chrome and Edge) shows the monitors as the computer arranges them; give each screen its monitor and it opens there, filling it. In other browsers drag each window where you want it once: it opens there next time.
• The layout is kept on the workstation for the position, so the next shift at the desk gets the same screens. With Offer to open these screens at sign-in, the full console offers them in one click after signing in.
• The browser asks to allow pop-ups (and, in Chrome and Edge, window management) the first time.

Disciplines

A screen limited to some disciplines shows their units, and the incidents that need them: an incident's own call type, the units sent to it, and the units its run card asks for (a crash with injuries is on the fire screen and on the police screen). The status bar shows the screen's disciplines as a coloured chip. Every screen still shows every emergency and every 911 call not yet on an incident.

Incidents and call entry screen

The queue of 911 calls and incidents, with the incident being worked beside it at full height, the new-incident form (F2) and the command line. It is the call taker's and dispatcher's working screen when the map and units are on other monitors. Everything works as on the full console (see The dispatcher screen).

Fire houses screen

Every station as a card, with every unit that lives there or is covering there.

• The bays: one door per unit. Shut and green, it is in quarters and available; open, the stripe is the colour of its status.
• The state: IN QUARTERS (all home), PARTIAL, COVERED (a move-up is in) or EMPTY (no unit available in the house, with how long).
• Tiles: each unit's status, its time in it, and what it is doing (its incident, or where it is covering). Click, drag onto an incident or right-click as on the unit board.
• Gaps: a kind of unit the house should keep in quarters but does not, with the best Move up in one click (see Move-ups).
• Across the top: houses with a unit in, empty houses, units out, and each kind of unit available out of its total, with every status in a bar.

Police status and mark-outs

Every police car in one table, with the map beside it: what it is doing, where, for how long, and when its status check is due.

• Doing: its incident (click to open it), or its mark-out: what it is doing on its own, such as a meal, court, fuel, a report or a follow-up.
• Where: the place given when it was marked out, its incident's address, or the nearest address to its AVL.
• Status check: a ring fills as the unit's time runs; amber near the end, red and flashing when the check is due, with how long past. A check is due after the status timer (Admin › Settings), or the mark-out's own time.
• OK records the status check (the unit is all right) and starts its timer again. MO marks it out; AV makes it available.
• The tiles across the top count available, on calls, marked out (by kind), out of service and checks due; click one to show only those.

From the command line

• MO 2A12 MEAL 315 Church Ave: mark out, with where. Without a place the unit's AVL position is kept. Add a note after a slash: MO 2A15 COURT 315 Church Ave / Courtroom 2. MO alone lists the agency's kinds.
• SC 2A12: status check, all right. Several units at once: SC 2A12 2A13.

Mark-outs are also on every unit's right-click menu and in the unit details, and a crew marks itself out from the mobile terminal (F7). A mark-out that can be pulled (a meal, a report) is offered by the recommendation when no available unit can fill a line: it is never sent without a dispatcher choosing it. The kinds, their times and which can be pulled are set in Admin › Settings.

Unit status screen

The unit board on its own, with larger tiles that also say what each unit is doing. Click, drag and right-click work as on the unit board of the full console.

Map screen

The dispatch map alone. It follows the incident or call selected on any of the position's screens, and draws marked-out units where they were marked out when their AVL is not fresh.

Incident and situation screen

The selected incident at full height (its calls, units, timeline, premises, recommendation and report), or, with nothing selected, the situation: what is waiting, status timers past due, move-ups and activity.

911 calls

Calls arrive from BitVoice's event stream, from EIDO pushed by any call-handling system, and from the ALI spill. The same call from several sources is merged into one line.

• Ringing calls show before they are answered, so the dispatcher sees what is coming.
• Location: the caller's position and uncertainty circle, the address when there is one, and STALE when the location is older than the setting in Admin › Settings. Rebid asks for a new one (BitVoice).
• + Incident on a call creates an incident from it, with the location and callback filled in. Calls already attached show their incident.
• Text to the caller, a video link and live video are on the call when BitVoice offers them.
• Its transcript (chapter 27) and what was heard, translation (chapter 28), the device location and floor (chapter 29), crash data (chapter 32), callers grouped (chapter 30) and non-emergency (chapter 31) calls are described in their own chapters.

Incidents

New incident: F2, the New button, + Incident on a call, right-click on the map (the address is filled in), a search result's New incident, or NEW SF 2001 Crystal Spring Ave on the command line. An address the recorded map knows exactly is placed on the map at once, so units are ranked by road straight away; anything else stays text to be placed.

The incident panel shows the recommendation, the units on it with their times, the caller and the timeline. Notes (N text) and priority notes (PN text) go on the timeline; priority notes are shown to every dispatcher and unit.

Close with a disposition (CL Handled); units still on it become available. Supervisors can reopen (RO #26-000123 reason).

Unit recommendations

For the selected incident bitdispatcher builds a recommendation from its call type's run card: for each line ("3 × engine", "1 × als"), the available units with that capability, closest first. Units already on the incident fill lines first, and scarce capabilities are filled before common ones so a multi-role unit is not spent where another would do.

• Drive times come from the recorded road network: posted or typical speeds, one-way streets obeyed, from the unit's fresh AVL position or its station. The reason says so: "1.7 mi by road, 4 min via Franklin Road".
• Outside the recorded map, or with no road between, the time is a straight-line estimate and is marked ~.
• A unit covering another station is ranked from where it is posted.
• Notes warn when no unit has a capability (consider mutual aid) or when sending the units would leave a station empty.

The recommended units are picked; Dispatch (Ctrl Enter, or REC) sends them. Click units to change the pick, Use recommendation to go back to it, or type D E1 M2. The ledger records what was recommended and what was sent.

Advisory suggestions

Beside the incident and on the Situation panel, the advisory panel may suggest something: two incidents that may be the same, a call that may be about an open incident, words that do not fit the call type ("smoke" on a fire alarm, "not breathing" on a fall), a repeat address, a priority 1 still without a unit, a unit on scene not heard from.

• Why: open a suggestion to see the evidence it relied on (with its source: unit, caller, dispatcher, machine), what it assumed absent, and the rule in plain words.
• Cannot determine is an answer: when sources of equal weight disagree ("smoke" and "no smoke" from two callers), it says so and you decide.
• Accept runs the suggested command as you (change the type, attach the call, dispatch the recommendation), or just notes it. Reject with a one-click reason; it is not shown again. Both are recorded.
• It advises only. It never merges, retypes, reprioritises or dispatches by itself, and core dispatch works the same when it is off or down.

Move-up and cover

When a station has none of a kind of unit it normally houses available (its engine is out on a call), the Situation panel shows a Move-up box with the move that fills the gap: the nearest unit of that kind by road, from a station that can spare one.

• Move E3 from ST3 posts the unit at the empty station in one click (MU E3 ST1). It stays available and is recommended from there.
• A move that would leave its own station empty is offered only after moves that empty nothing, and says "empties ST4".
• When the station's own unit is back, a Home button sends the cover home (RET E3).
• Which kinds are kept covered is set per unit type in Admin › Settings (front-line engines, ladders, medics, ambulances and patrol by default).

Unit statuses

Each status has a timer (Admin › Settings); a unit past its timer turns amber on the board and is listed under Status timers past due. Statuses are given on the command line (ER E1, AV E1 M2), from the unit's tile, or by the crew on the mobile terminal (chapter 20). AVL positions older than the stale setting are drawn dashed and not used to rank units.

The dispatch map

The map is drawn from bitdispatcher's own recorded map (chapter 46): roads by class with route shields and one-way arrows, county and town lines, house numbers and hydrants at street level, fire, police, EMS and hospitals in their own colours.

• Search (top right): an address (2001 Crystal Spring Ave), an intersection (Orange & Williamson, and, / or at), a street, or a place. Arrow keys and Enter; New incident from a result. A house number not on file offers the nearest numbers on the same street.
• Right-click anywhere: what is here (the nearest address or road) and New incident here.
• Routes: for the selected incident, the road route of each recommended, chosen and assigned unit, with its arrival time on the unit.
• Day / Night switches the palette; Photos (F7) puts aerial photos under the streets; Fit all shows every unit and incident; OOS and Stations show or hide those.
• The dashed cyan line is the edge of the recorded map: beyond it the online map is shown, muted, and drive times are estimates.

Aerial photos

Dispatchers find places by rooftops ("the white house behind the church"). With Photos on (F7 or the button on the map), aerial photos are the base and the streets, names and house numbers are drawn over them.

• Photos are recorded in Admin › Map: full detail near the centre, less detail out to the edge. Zooming in past what was recorded enlarges the photo below it, so the map never goes blank.
• The chip beside the button says when the photo at the map centre was taken, where the source publishes it ("2023, 12 inch ortho"): an old photo is never taken for today.
• Where nothing is recorded the server fetches the photo when asked and keeps it, unless that is switched off.

The command line

Type a letter anywhere on the dispatcher screen and the command line takes it (or press /). Tab completes, arrow keys recall earlier commands, results stay on screen. HELP lists every verb.

Keyboard

Mobile data terminal

A crew signs in on the vehicle's laptop or tablet with a field account (Admin › Users) and the unit's call sign as the position (E1). The terminal opens at once.

• The assignment: type, priority, address in large type, the road route drawn on the map with its time and the roads it takes, the callback, the other units on it, and the notes from dispatch, newest first. Priority notes stand out.
• Status buttons: one tap each, with the usual next status highlighted; F1 to F6 on a keyboard.
• Notes to dispatch go on the incident's timeline as the unit's.
• GPS: the device's position goes to bitdispatcher as AVL every few seconds (allow location when the browser asks).
• Emergency: hold the button for two seconds. Dispatch sees the emergency banner and acknowledges it.
• Photos on the map, as on the console.

A crew acts only as its own unit. Dispatchers can open the terminal at #/mdt to see what a unit sees (the position names the unit).

Incident reports (RMS)

Officers and dispatchers write incident reports at #/rms, or with Report on an incident (the incident's location and time are filled in). A crew on a mobile terminal writes its own from the car.

• Sections: when and where (location type, whether it was cleared), offences, people, property, vehicles and the narrative. Codes are shown in words and picked in one click; the screen asks only for what the offence needs (a weapon for an assault, the method of entry for a burglary, the drug for a seizure, the injury and the relationship to each suspect for a crime against a person).
• Saved as you type: a draft saves itself a moment after typing stops.
• Before it is submitted lists what is still missing for incident-based reporting: the FBI's NIBRS data elements and edits, with Virginia's additions (property quantity, the licence plate reader indicator, drug types). Submit is available once the list is empty.
• Approval: a supervisor other than the author approves it, or returns it with what to fix. Every save and decision is a ledger event.

The state file

Supervisors see State IBR file under the list. Choose the month and press File: bitdispatcher writes the month's approved reports in Virginia's incident-based reporting format (the FBI's position-based flat file with Virginia's additions, 2025.2 specification): Group A incidents with their offence, property, victim, offender and arrestee segments, Group B arrests, or a zero report when there is nothing. Upload it to the Virginia Crime Repository by the 15th of the next month. Reports that fail the checks are left out and listed. Making the file is a ledger event, with its SHA-256.

NCIC inquiries

NCIC (on the reports screen, F9 on the dispatcher screen, or on the mobile terminal) sends an NCIC 2000 inquiry through the state's message switch: wanted person (QW: name with date of birth, or name, sex and race), vehicle (QV: plate and state, or VIN), gun (QG) and article (QA) by serial.

• Each inquiry needs a purpose code and a reason (the incident number): CJIS requires both.
• The ledger keeps who asked, when, the purpose, the reason and whether a record came back. The terms are kept only as a fingerprint and the answer not at all.
• A hit is a lead: confirm it with the entering agency before any action.
• Administrators set the agency's ORI and the switch in Admin › Settings: off, the training simulator (plates, names and serials starting with TEST return a record; every answer says SIMULATED), or a switch adapter by TCP.

The supervisor board

The whole picture at a glance, for a supervisor's second monitor or a wall display. It only shows: nothing on it changes anything.

• Across the top: 911 calls not yet on an incident (red when one is ringing), incidents waiting for a unit with their priorities and the oldest wait, active incidents, units available out of those in service, units marked out by kind, status checks due, and empty stations.
• Where the work is: the map shaded by how much is open in each area (open incidents plus twice those waiting), with each area's open, waiting and available units. Areas are the county's emergency service zones or beats when they are loaded (Admin › County GIS); otherwise each station's area, the part of the map nearest it. An area with incidents waiting and no unit free, or two waiting, is SWAMPED.
• By discipline: a ring of every status with the number free, and each kind of unit available out of its total.
• Busiest areas, worst first, and new incidents over the last six hours by priority.
• People: each signed-in position's workload and breaks, and the care list after critical calls. See Breaks and workload (chapter 37) and Critical calls and care (chapter 36). A possible line attack shows across the top (TDoS (chapter 41)).

Choose the disciplines across the top, or open it from Screens with them. Full screen is beside the clock.

Operations report

Admin › Operations report builds the Daily Operations Report and shift debrief for a period from the ledger alone: nobody retypes what CAD already holds.

• Periods: this shift, last shift (from the shift change times in Settings), today, yesterday, last 7 days, or any range.
• Contents: 911 calls, incidents, dispatches and closes; incidents by hour, type, priority and disposition; response times (call to dispatch, dispatch to en route, call to on scene, on scene to clear) as median and 90th percentile; priority 1 incidents; incidents open at the end (the turnover); status timers run out, once per episode; emergency activations and how fast they were acknowledged; each person's actions; unit demand by phase and utilisation; system and configuration changes.
• Drill-down: click any number to see the ledger events it is made of.
• Certify: once the period has ended, a supervisor certifies it with optional context. The certification is a ledger event with the report's hash; if anything changes later, the report says it changed since it was certified.

The advisory codebook

Admin › Advisory panel holds the codebook: the rules the panel uses, in plain words, with their thresholds.

• Mode: off; shadow (evaluated and kept in the ledger, shown to supervisors only, so the rules can be judged before anyone relies on them); live (dispatchers see suggestions). A new system starts in shadow.
• Source precedence: drag to order unit, caller, dispatcher and machine evidence, strongest first. A stronger source outweighs a weaker one; equal sources that disagree give "cannot determine". A silent alarm panel never outweighs a person who can see smoke.
• Rules: switch each on or off, set its severity, distances, times, words and the call types they already fit, and edit its explanation.
• Test replays the recorded ledger for the last day, week, month or quarter with the rules on screen, saved or not, and counts what each would have suggested.
• Save version keeps the change with a note; every version is listed and can be restored in one click. Decisions (accepted, rejected and why) are listed: rejecting a suggestion is never treated as an error.

Agency scripts

JavaScript script and Lua script add a rule that is the agency's own code, versioned, tested and switched between shadow and live with the rest of the codebook. A script defines one function for its hook:

An incident script sees the number, type and its name, priority, status, address and position, minutes since it was created, notes, units on it, hazards at its premises, caller name and number of calls. Try it runs the script once on the newest real incident (or report) and shows what it returned, what it logged with log(), and how long it took; show the input shows exactly what it saw.

• A script has no files, network, process, environment or code loading: only its input and log().
• It always has a time limit (50 ms unless set, at most 1 s). It runs after the CAD's state is released, so a slow script never holds dispatch up.
• A script that fails or runs out of time is shown under its rule and in the test; the panel and the recommendation carry on without it and say so.

Ledger and audit

Every action that changes state, takes attention or creates accountability is an event in an append-only ledger: who (user, role, position, session), when, what, on which incident, unit or call. Each event carries the hash of the one before it, so a change anywhere breaks the chain.

• Verify checks the whole chain and reports the first broken link, if any.
• Search by type, person or text; events are newest first.
• The current state is rebuilt from the ledger at every start, and reports are built from it.

Transcripts and what was heard

When call handling transcribes calls, each call shows its transcript live, the caller in white and the call-taker in blue, and Heard on the call: the details pulled out of it, so the call-taker types less.

• What is pulled out: the address, weapons, vehicles, plates, the direction someone went, how many people, injuries (not breathing, unconscious, bleeding), hazards (fire, trapped, gas, a dog, an electric car) and the caller's name. Urgent ones are red and come first, and show on the call in the queue.
• Use: on an address, places the call's incident there (a map search, exact when the map knows the number; check it on the map). Before there is an incident it copies the address.
• Where transcripts come from: BitVoice's live transcript when it sends one, its transcript of a recorded call afterwards, or any speech-to-text engine posting to the transcript hook (Hooks (chapter 42)).
• Rules, not a model, pull out the details: the same words always give the same details, and each shows the words it came from.

Callers in another language

A caller speaking Spanish is translated line by line on the transcript, marked Spanish · translated; Show original shows what was said.

• The phrasebook built into bitdispatcher translates the words of emergencies between Spanish and English, offline. When it knows too few of a line's words it says gist only: get an interpreter.
• Say it in Spanish: press a ready phrase (Help is on the way, Is the person breathing?, Get out of the house now…) or type in English; the translation shows as you type. It is kept on the transcript, texted when also text it is ticked (BitVoice), and the speaker button says it aloud on this console.
• A language model (optional, Admin › Integrations) translates what the phrasebook cannot, on an OpenAI-compatible chat API: one on your own server, or a service under your agency's agreement. What callers say is sent to it, so it is off until an administrator switches it on.

Where the caller is: device location and floor

Each call says how its position was found: device location (the phone's own fix, often a few metres, the FCC's dispatchable location), GPS, cell sector or on file.

• Floor: a civic floor from call handling, and the altitude of a 3-D fix with its uncertainty, which tells which floor of a tall building.
• Trail: a caller who moves leaves a dotted trail on the map of every position received, the selected call's brighter.
• Rebid asks for a new position; STALE marks an old one (Admin › Settings).

Many callers, one incident

When several people call about the same thing, the queue groups them under N callers about one place: calls near each other (300 m by default) within a short time (10 minutes).

• Make the incident from one call; the group then offers Attach N to that incident, one press for the rest.
• With an incident selected, the group offers to attach all of them to it.
• Calls already on different incidents are never grouped together. Change the distance and time in Admin › Settings (chapter 48).

Non-emergency calls

Calls on a non-emergency line, or whose caller says something that is seldom an emergency (a noise complaint, parking, a power outage), go to the Non-emergency list under the 911 calls, with the reason.

• Anything urgent said on the call (a weapon, not breathing, fire) keeps it an emergency, whatever else was said; so does crash data.
• It is an emergency moves it back; a person's decision always stands and is recorded.
• Send the usual reply texts the reply set in Settings, such as where to report online.
• Set the lines, the words and the reply in Admin › Settings (chapter 48).

Vehicle crash data

A vehicle's automatic crash notification arrives as a call with CRASH DATA: the vehicle, delta-V (the change of speed in the crash), the direction of the impact, rollover, airbags, occupants and belts, and its power (electric, hybrid).

• Injury risk is high for delta-V of 40 km/h (25 mph) or more, a rollover, or 20 km/h with no belt, several impacts or a side impact; medium for 20 km/h or any airbag. Suggestions follow: ALS, extrication, air medical, more than one ambulance, and for an electric car a word to the fire crews.
• It is a rule of thumb from field-triage criteria to help pick the response, not a diagnosis. Change the thresholds in Admin › Settings (chapter 48).
• Crash notifications reach bitdispatcher through the crash hook (Hooks (chapter 42)).

Drones

With a drone system connected (Admin › Integrations (chapter 49)), an incident on the map shows Send a drone. bitdispatcher sends the drone system the incident, its type, priority, position and summary; the drone system decides which drone flies.

• Launch rule: incidents of call types marked Send a drone (Call types) ask by themselves, once each.
• The drone reports back to the drone hook: its status, position, battery and live picture. It flies on the map, its status goes on the incident's timeline, and Live picture opens its video.

Radio transcripts

When BitVoice or the radio system transcribes transmissions, a unit's words go on its incident's timeline in italics (E1: on scene, smoke showing), so the record holds what was said on the air. The words come with BitVoice's radio events, or through the radio-words hook (Hooks (chapter 42)).

Incident summaries

Every open incident has a Summary: what and where, what the callers said that matters (the urgent first), the priority notes, the hazards on file at the place, who is on scene and who is responding, and a drone over it.

• It is built from the record by rules, so it says nothing the record does not; it changes as the incident does.
• It shows on the incident panel (Copy puts it on the clipboard) and at the top of the crews' mobile terminal (chapter 20).

Critical calls and care

Some calls stay with the people who take them. An incident is a critical call when its call type is marked critical (cardiac arrest, shooting and stabbing to begin with) or when words such as not breathing, CPR, officer down or infant are said on it.

• Everyone who worked it (the call-taker and the dispatchers who entered, sent or noted) goes on the care list on the supervisor board (chapter 23), with the reason.
• The supervisor offers a break to each person, arranges a debrief (with a note of when and where) and marks it Done. Each is recorded.
• Once the incident closes, each person sees a quiet line on their own console: that was a hard call, take a moment, and the support line set in Settings. Take a break puts them on a break.
• Set the words, the support line and the call types in Admin › Settings (chapter 48) and Call types (chapter 45).

Breaks and workload

The cup on the status bar puts you on a break and back; the supervisor board shows it.

The board's People panel lists every signed-in position: calls in the last hour, actions, time since the last break, critical calls in four hours, and a state:

• watch: two hours without a break (Settings), a critical call, or a very busy hour;
• relieve: three hours without a break, or two critical calls.

It names the least busy positions free to relieve someone.

Training

Tick Training when signing in. The console opens a board of its own, marked TRAINING in violet: the agency's units, stations, call types, premises and map, on a ledger of its own. Each trainee has their own board, even on a shared account; one left idle for two hours is cleared. Nothing done there reaches the live board, and administration is closed.

• Scenarios play calls, transcripts, crash data and radio traffic into it on the agency's own map: a house fire with three callers, a cardiac arrest in Spanish, an interstate crash with crash data, an armed robbery, and a flood of calls on the line.
• Score checks what was done: every caller on an incident, the call type, the units sent, a reply in the caller's language, and each incident's quality review (chapter 39).
• Reset empties your board and starts it again; other trainees' boards are untouched.

Quality review

Every incident checked against the centre's standards:

Each incident gets a score over the checks that applied; the page lists them for a period, each call-taker's average and each check's pass rate. Press an incident for its checks. A review is a coaching aid: a missed question may simply not be on the transcript.

Working through an outage

Each console keeps the last board it had. When the server cannot be reached, an amber bar says so and the console shows that board, as it was, with the time. A console reloaded during the outage still opens.

• Offline log: new incidents, notes and unit statuses are kept in this browser, in order.
• When the server is back the bar offers Send them now: each goes in marked "entered offline at 14:02 by jlee"; anything the server refuses stays, with the reason.
• For a server that never goes away, run bitdispatcher at two sites (Several sites (chapter 52)): each keeps working when the other is cut off.

Line attacks (TDoS)

A telephony denial-of-service fills the lines so real callers cannot get through. bitdispatcher watches the calls and raises Possible telephony denial-of-service on every console and the supervisor board when:

• far more calls arrive in a minute than usual (five times the last hour's rate, at least 12, or the number set in Settings);
• one number calls five times or more in ten minutes (shown as ×5 on its calls);
• eight or more silent or very short calls arrive in five minutes.

A supervisor acknowledges it; it clears after ten quiet minutes. Keep answering: a real emergency can be among them.

Hooks and connections

Other systems send to bitdispatcher with Authorization: Bearer and the hook token set in Admin › Integrations (chapter 49). Without a token the hooks are off.

The drone system's launch address and token, the launch rule, and the optional translation model are set on the same screen.

Units

Each unit has a call sign (E1, M12, 2A31), a type (which gives its capabilities), a station, and optionally extra capabilities, the radio's unit ID and alias (so radio emergencies and positions find it) and an agency for mutual aid. New units start out of service.

Stations

Stations are where units are based. Their positions place units with no AVL, are where move-ups post covering units, and are where drive times start from when a unit's AVL is stale. Type the address and pick it from the recorded map's suggestions: the position fills in. Or type the position.

Call types and run cards

A call type is a kind of incident (SF structure fire, CHEST chest pain) with a default priority, a discipline and a run card: how many units of each capability it needs ("3 × engine, 1 × ladder, 1 × als"). Recommendations fill the run card. Capabilities come from unit types (Admin › Settings) and units.

Critical call puts the people who work one on the supervisor's care list (chapter 36); Send a drone has the drone launch rule ask for a drone (chapter 33).

Recording the map

bitdispatcher records the OpenStreetMap data for a circle around your centre into its own store, then draws the dispatch map, routes units by road and finds addresses from that store with no internet.

1. Choose the centre (a preset, the console map centre, or a position) and the radius (100 miles covers a county and its mutual-aid neighbours).
2. The downloads the area needs are listed at once (state extracts from Geofabrik). Press Record.
3. Recording takes a few minutes; dispatching carries on with the current map, which stays until the new one is complete. Each recording is a ledger event.

Bring up to date applies OpenStreetMap's daily change files (Geofabrik publishes one a day per extract) since the map was recorded: a few megabytes instead of the whole extracts, then the store is built again. Record again for a new area, or when the map is more than a year behind; downloads already made are checked against their published checksums and reused.

Aerial photos

Below the map, Aerial photos records photos around the same centre: full detail within the radius you choose (zoom 19: rooftops and cars) and less out to the edge (zoom 16). The estimate (tiles, size, time) follows the form; tiles already recorded are skipped, so a stopped recording carries on where it was. The sources and their terms are listed: VGIN's Virginia Base Mapping Program for Virginia, USGS The National Map (NAIP, public domain) elsewhere.

County GIS

The county's own GIS is more exact than OpenStreetMap where the county keeps it. Admin › County GIS imports its layers for the recorded map's area into a store of their own, kept when the map is recorded again.

1. Paste the layer's ArcGIS REST address (…/FeatureServer/0), choose a preset (Virginia's statewide address points and road centrelines from VGIN, or Overture's open addresses), or upload a GeoJSON file or a CSV with latitude and longitude columns. The layer is read at once: its name, how many records lie in the area, and a guess at the kind and the fields.
2. Check the field mapping against the sample values shown beside each field, and press Add layer.
3. Import now reads every layer that is on. The store in use stays until the new one is complete; each import is a ledger event.

Search results from county data are marked as such. ArcGIS layers are read with https only, in batches, so services that cap a query still give everything.

Settings

• Agency name and incident number prefix.
• Status timers: minutes in each status before the dispatcher is prompted for a status check.
• Mark-outs: the kinds a unit can be marked out for (code, name, status-check minutes, which units, whether it can be pulled for a call). See Police status and mark-outs.
• Stale ages for AVL and caller locations.
• Shift changes (07:00 19:00) for the operations report's shift periods.
• Travel estimates outside the recorded map: road factor and average speed.
• Calls: how near and how soon callers are grouped (chapter 30); the non-emergency (chapter 31) lines, words and reply.
• People: the critical-call (chapter 36) words, the support line, and when to suggest a break (chapter 37).
• Quality review standards and the questions to listen for (Quality review (chapter 39)); crash data thresholds (Vehicle crash data (chapter 32)); the line-attack (chapter 41) warning level.
• Map beyond the recorded area: the online tile source (fetched and cached by the server, never by browsers), and the map's starting centre and zoom.
• Unit types: code, name, discipline, capabilities, and whether stations keep them covered (move-up suggestions).
• Dispositions an incident can be closed with.

Integrations

How bitdispatcher connects to call handling and radio. Each integration shows its live state and a checklist of the hooks in use.

• BitVoice: its address and API token, and its certificate pinned with Read it. See BitVoice (chapter 50).
• EIDO: the token a call-handling system uses to push EIDO to /api/ng911/eido.
• Radio: the secret for radio webhooks to /api/radio/events.
• ALI spill: the TCP address bitdispatcher listens on, and Test.
• Hooks, drones and translation: the hook token for transcripts, crash data, drone reports and radio words, the drone system, and the optional language model. See Hooks and connections (chapter 42).

BitVoice

bitdispatcher uses every hook BitVoice offers a CAD, and needs none of them: without BitVoice, ALI and EIDO from any call-handling system work the same way.

Delivery failures BitVoice reports show on the Situation panel.

Users and positions

Passwords are generated or at least 12 characters. Five wrong passwords lock an account for 15 minutes. The positions list shows who is signed in where.

A public demonstration account. Tick Public demonstration account on a viewer, dispatcher or supervisor and the sign-in page offers Sign in as the demo user: one press signs a visitor in, with no password, at a position of their own (WEB- and four letters) unless they typed one. One account at most, never an administrator or a field crew, and wrong passwords typed for it do not lock the button out. Use it only on a demonstration server: anyone who can reach the server can then work as that user.

Several sites

bitdispatcher runs at two or more sites, all live at once: a main centre and a backup centre, or neighbouring counties sharing a system. There is no primary to fail over from.

• Each site writes its own ledger and copies the others' over https, every second, each event checked against its site's hash chain before it is kept.
• The same picture everywhere: every site applies all events in the same order (by time, then site), so sites that have the same events show the same state, whatever order they arrived in.
• Cut off, a site keeps working on its own; when the link returns the events are exchanged and both sites settle on the same state. If two sites changed the same unit meanwhile, the later change wins, at both.
• Incident numbers carry each site's prefix (N26-000001, S26-000001), so two sites never give the same number.
• Setting up: give this site a name and prefix and save; copy its token and certificate fingerprint to the other sites. Add each other site with its address, its token and its fingerprint (Read it reads the certificate; check it against the one shown on that site). The status bar shows how many sites are linked.

Premises and hazards

A premise is what the centre knows about a place before anyone calls: hazards to the crews (a dog, oxygen in use, an armed occupant, a Knox box), notes and contacts. It is shown with every incident within 40 m of it, on the dispatcher's incident panel and on the mobile terminal, hazards first in red. An expired premise (a temporary hazard past its date) is shown as expired.

• Add a premise: type the address and pick it from the suggestions (that places it), click the common hazards or type others, add notes and contacts.
• Search by address, name, hazard or id; click one to change or remove it.
• Supervisors and administrators change premises; everyone sees them. Each change is a ledger event.

Premises from an earlier system come in through Migration.

Migration from an earlier CAD

Admin › Migration brings an earlier CAD system's data over from the spreadsheets it exports, CSV or tab-separated (UTF-8 or Windows encoding).

1. Drop the file on the page. Its columns are matched to ours from their names, and every row is checked at once.
2. Choose what it holds if the guess is wrong: premises and hazards, incident history, or call types and run cards. Change any column in the table; the first row's value is shown beside it.
3. Import the rows that passed. The rows that did not are listed with why (no position and an address the map does not know, a date it cannot read).

Importing the same file again replaces what it brought (premises by their id, call types by their code); history is added again, so import it once. Address points go to County GIS, which reads CSV files with latitude and longitude as well.

This manual

This manual is built into bitdispatcher, so it is there wherever it runs, with no internet. Every screen's Help link opens its chapter; the palette (Ctrl K) finds chapters by name. The same chapters are printed as docs/manual/index.html and a PDF by make manual.