| `name` | String | required | Internal provider identifier referenced by `provider://<name>` URLs. Must be unique within the provider list. |
| `urls` | List | required | Ordered failover URL list for this provider. Tuliprox rotates through these URLs within a request when failover is triggered. |
| `provider_url_selection_policy` | Enum | `resume_last_working` | Controls how a new request chooses its starting URL. `resume_last_working` starts at the last successful URL. `restart_from_first` always begins again at `urls[0]` and only fails over within that request. |
| `dns` | Object | unset | Optional DNS/IP rotation settings for the provider. See the table below. |
| `refresh_secs` | Int | `300` | The interval in seconds the background task resolves the hostnames. (Minimum effective value is 10). |
| `prefer` | Enum | `system` | Which IP protocol to prefer during DNS resolution. Options: `system`, `ipv4`, `ipv6`. |
| `max_addrs` | Int | `None` | Hard limit on the number of resolved IPs to retain per host. |
| `schemes` | List | `[http, https]` | The HTTP schemes that IP connection rotation applies to. |
| `keep_vhost` | Bool | `false` | If `true`, the `Host` header retains the original `hostname[:port]`. If `false`, it uses `IP[:port]`. Essential for reverse proxies upstream! |
| `on_connect_error` | Enum | `try_next_ip` | Policy on TCP connection failure. Options: `try_next_ip` (cycles to the next resolved IP for the same host), `rotate_provider_url` (instantly fails over to the next URL in the `urls` list). |
### 1.1 DNS Resolved IP Persistence
Resolved IPs are persisted to `{storage_dir}/provider_dns_resolved.json` (not to `source.yml`).
This file is written atomically after each DNS refresh cycle and read at startup to seed DNS caches before the
background resolver
completes its first cycle.
On config hot-reloads, DNS caches are carried over from previous provider instances so that resolved IPs are available
| `name` | String | Yes | | Internal reference ID for Tuliprox. Must be strictly unique. Critical for persistent UUID generation! |
| `type` | Enum | No | `m3u` | Allowed: `m3u`, `xtream`, `stalker`, `library`, `staged`, `emby`, `jellyfin`, `plex`, and `m3u_batch` / `xtream_batch` / `stalker_batch` (CSV offloading). Stalker inputs use the portal handshake/catalog flow instead of a plain playlist download. |
| `url` | String | Yes | | The Provider URL. Tuliprox supports magic scheme prefixes: `http(s)://`, `file://`, `batch://`, and **`provider://my_failover_provider`** (for the Failover System above). |
| `username` / `password` | String | Often | | Mandatory for `xtream` and for Stalker inputs using `credentials_only` or `mac_plus_credentials`. Stalker inputs using `mac_only` do not need them. |
| `enabled` | Bool | No | `true` | If `false`, this input is completely ignored in all processing. |
| `sequential_group` | Int | No | | Optional non-zero process-wide group ID. With `process_parallel: true`, complete refreshes of inputs sharing an ID run one after another, including all Stalker page slices. Different groups and ungrouped inputs may overlap. Not valid for `staged` inputs; a staged overlay inherits its provider job. |
| `cache_duration` | String | No | `0` | **Crucial:** Determines how often Tuliprox actually downloads the raw list from the provider. At `1d` (1 day), Tuliprox serves from its local `.db` for 24 hours, even if you trigger hourly updates. This heavily protects against provider bans! Supported units are `s`, `m`, `h`, and `d`. If `cache_duration` is set, the cached provider playlist stored on disk is reused for subsequent updates instead of downloading it again. |
| `persist` | String | No | | Optional path template (e.g., `./playlist_{}.m3u`) to permanently store the downloaded raw provider list locally on your disk. The `{}` in the filename is filled with the current timestamp. For `m3u` use a full filename. For `xtream` use a prefix like `./playlist_`. |
| `method` | Enum | No | `GET` | HTTP Request method for playlist downloads (`GET` or `POST`). |
| `exp_date` | Mixed | No | | Expiration date as `"YYYY-MM-DD HH:MM:SS"` or Unix timestamp. Used for status tracking and Panel API logic. |
| `headers` | Dict | No | | Custom HTTP headers for the download (e.g., `User-Agent: My-Player`). |
| `epg` | Object | No | | Allows mapping of external XMLTV files (see [below](#input-subsections-object-keys)). |
| `aliases` | List | No | | Connection pooling / Sub-accounts (see [below](#input-subsections-object-keys)). |
| `staged` | Object | No | | Staged overlay settings. Only valid when `type: staged` (see [below](#input-subsections-object-keys)). |
| `panel_api` | Object | No | | Automated reseller account generation (see [below](#input-subsections-object-keys)). |
#### Minimal Stalker Input Example
```yaml
inputs:
- name:stalker_main
type:stalker
url:http://portal.example.com/c/
stalker:
auth_mode:mac_only
mag_preset:generic_safe
endpoint_preference:auto
device:
mac_address:'00:1A:79:12:34:56'
enabled:true
options:
stalker_pre_resolve_playback:false
stalker_runtime_resolve_playback:true
```
Stalker refreshes write pages into an unpublished generation. Live, VOD, series, and EPG selected for one refresh
become active together only after the complete selection is durable. Until then, Tuliprox continues serving the
previous complete snapshot; a first import exposes no partial catalog. A saved checkpoint resumes after restart.
When `process_parallel` is enabled, progress messages include the input name. Targets wait for every enabled input in
their source and begin as soon as that source is ready, without waiting for unrelated sources.
Use `stalker_pre_resolve_playback: true` if you want Tuliprox to materialize playback URLs during refresh whenever the portal already grants them.
Keep `stalker_runtime_resolve_playback: true` when the portal uses expiring or session-bound temp links that may need a fresh `create_link`
call later during playback.
Stalker supports four authentication modes:
*`auto`: requires either a MAC address or a complete username/password pair.
*`mac_only`: requires `stalker.device.mac_address`; username/password are ignored.
*`credentials_only`: requires username and password; no MAC address is required.
*`mac_plus_credentials`: requires both a MAC address and username/password.
| **`http(s)://`** | Remote Server | Standard method for downloading playlists from provider endpoints. |
| **`file://`** | Local Storage | Reads a playlist directly from the host filesystem. Useful for manual backups or pre-processed files. |
| **`provider://`** | Failover System | Resolves the URL via internal `provider` definitions. **Pro-Tip:** Use this to implement automatic rotation or failover between multiple mirrors/gateways of the same provider. New requests honor `provider_url_selection_policy`, so they can either resume from the last healthy URL or always restart from the first URL. |
| **`batch://`** | CSV File | Dedicated scheme for bulk alias management. Points to a local `;` separated CSV file (e.g., `batch://./aliases.csv`). |
| `skip_live` / `skip_vod` / `skip_series` | Bool | `false` | Immediately ignores entire categories during Xtream or Stalker ingestion. Saves massive amounts of RAM and runtime if you only want specific clusters from a provider. |
| `xtream_live_stream_use_prefix` | Bool | `true` | Injects the `/live/` prefix into URLs. |
| `disable_hls_streaming` | Bool | `false` | Forces Tuliprox to play Live-TV as a raw MPEG-TS (`.ts`) stream, skipping HLS (`.m3u8`) reverse-proxy handling, and forcing direct TS endpoints. |
| `resolve_tmdb` | Bool | `false` | Enables TMDB queries for this specific input based on parsed titles to fill missing posters and release years. |
| `resolve_background` | Bool | `true` | Metadata scans run asynchronously in the background so the general playlist update (which blocks clients) finishes instantly. |
| `resolve_series` / `resolve_vod` | Bool | `false` | Fetches missing details like Plot or Cast via the Provider's API (`get_vod_info` / `get_series_info`). |
| `probe_series` / `probe_vod` | Bool | `false` | Allows explicit FFprobe analysis of movies or entire TV show seasons. |
| `probe_live` | Bool | `false` | Allows FFprobe to periodically tap into Live-TV streams in the background. |
| `probe_live_interval_hours` | Int | `120` | Interval after which a Live stream is re-analyzed (Important as backup streams often change resolutions). |
| `resolve_delay` / `probe_delay` | Int | `2` | **Ban Protection:** Hard wait time (in seconds) between API or Probe requests to the *same* provider! Prevents API spamming. |
| `resolve_filter` | String | - | Filter expression to selectively resolve only entries matching the condition. Uses the same Filter syntax. |
| `probe_filter` | String | - | Filter expression to selectively probe only entries matching the condition. Uses the same Filter syntax. |
| `stalker_pre_resolve_playback` | Bool | `false` | Stalker-only: resolves `create_link` during playlist processing and persists the returned playback URL when the portal already grants one. Useful when you want the playlist/export step to materialize playable stream URLs up front. |
| `stalker_runtime_resolve_playback` | Bool | `false` | Stalker-only: lets the reverse-proxy retry `create_link` during playback when a persisted Stalker URL has gone stale or is rejected by the portal. This is the recovery path for temp links and expired session-bound stream URLs. |
* Stalker playlist preview in the Web UI now works through the same protected playlist endpoints used for other input types.
*`stalker_pre_resolve_playback` and `stalker_runtime_resolve_playback` are complementary:
*`stalker_pre_resolve_playback: true` tries to turn portal `cmd` values into concrete playback URLs during refresh.
*`stalker_runtime_resolve_playback: true` retries `create_link` later if the stored playback URL is stale, temp-link based, or rejected after processing.
* Temp-link variants (`nginx_secure_link`, `flussonic_tmp_link`, `wowza_tmp_link`) are persisted as explicit playback modes
and reused during runtime refresh, instead of being flattened into a generic direct-URL path.
* When pre-resolve does not materialize a URL, Tuliprox keeps the Stalker item metadata and playback descriptor but does
not leak the raw `cmd` into the exported playlist URL field.
* If pre-resolve is disabled or the portal refuses to resolve a specific item during refresh, the item can still remain playable
later through runtime resolution, assuming the reverse-proxy path is used and runtime resolve is enabled.
* Runtime refresh reuses a cached Stalker client per input configuration and treats the session TTL as a soft re-handshake boundary.
If refresh still cannot resolve a playable URL, Tuliprox invalidates the stale persisted URL instead of continuing to serve it indefinitely.
* Stalker EPG import now also consumes the portal bulk-EPG endpoint during processing when Stalker playback pre-resolve is enabled.
* The bulk-EPG path is streamed and batch-persisted to reduce peak memory pressure on large portals, but portal-specific
tuning for pathological datasets is still a separate follow-up topic.
* Supported Stalker playback transports are currently `http` and `https` only. `rtmp://` / `rtsp://` commands are rejected
explicitly because Tuliprox's reverse-proxy path does not relay those schemes.
* Fresh temp-link resolution is implemented. The still-open edge case is whether a specific portal also requires extra forwarded
cookies or headers on the final media request after temp-link resolution.
| **`type`** | Enum | No | `xmltv` | Defines the source format. Supported values are `xmltv` and `ics`. Existing configurations without `type` are treated as `xmltv`. |
| **`url`** | String | Yes | | The EPG source URL or local path. For XMLTV sources, use **`auto`** with Xtream inputs to automatically generate the native XMLTV URL using your credentials. For ICS sources, this points to an `.ics` calendar file. |
| **`priority`** | Int | No | `0` | Determines the lookup order. **Lower numbers have higher priority.** For example, `-2` is processed before `0`. Use negative numbers for primary sources. |
| **`logo_override`** | Bool | No | `false` | If set to `true`, channel logos from the provider are replaced by icons found in the EPG source. This mainly applies to XMLTV sources. |
| **`channel_id`** | String | ICS only | | Required for `type: ics`. Defines the generated XMLTV channel ID for the imported calendar. Playlist entries can reference this value through their EPG ID or receive it through Smart Match. |
| **`channel_title`** | String | No | | Optional display name for the generated ICS EPG channel. If omitted, `channel_id` is used as fallback. This value is also used as a Smart Match candidate. |
| **`match_names`** | List | No | `[]` | Optional additional names for matching the generated ICS channel to playlist entries. Useful when the playlist channel name differs from the calendar title, for example `F1`, `Formula One`, or `Formel 1`. |
| **`ics`** | Object | ICS only | | Additional configuration for `type: ics`. See [ICS Calendar Source Parameters](#ics-calendar-source-parameters-ics). |
#### XMLTV Sources
XMLTV is the default source type. Existing configurations remain valid and do not need to be changed.
```yaml
epg:
sources:
- url:"auto"
priority:-2
logo_override:true
- url:"https://example.org/xmltv.xml"
priority:0
```
The following configuration is equivalent:
```yaml
epg:
sources:
- type:xmltv
url:"https://example.org/xmltv.xml"
priority:0
```
For Xtream inputs, `url: "auto"` automatically generates the provider's native XMLTV endpoint from the configured
input URL, username, and password.
The ICS-only fields `channel_id`, `channel_title`, `match_names`, and `ics` are rejected on `type: xmltv` sources. This
keeps accidental calendar settings from changing XMLTV download or runtime behavior.
#### ICS Calendar Sources
ICS sources import iCalendar files and convert their `VEVENT` entries into XMLTV-compatible programme entries.
Recurring events using `RRULE`, `RDATE`, or `EXDATE` are detected but not expanded yet; Tuliprox imports the base
`DTSTART`/`DTEND` occurrence and logs one aggregated warning per parsed source when such entries are present.
An ICS source creates exactly one virtual EPG channel. It does **not** create a playlist channel or stream. The generated
`channel_id` must therefore be assigned to an existing playlist entry by one of the following mechanisms:
* the playlist entry already uses the same EPG ID,
* a mapper rule assigns the EPG ID,
* Smart Match matches the playlist channel name against `channel_id`, `channel_title`, or `match_names`.
The generated programmes are written into the regular XMLTV output. M3U and Xtream use the same internal
`epg_channel_id` assignment, so the same `channel_id` works for both input types. If an Xtream provider supplies a valid
but unwanted EPG ID, use a mapping rule to overwrite the stream's EPG ID with the ICS `channel_id`.
| **`timezone`** | String | No | `UTC` | Fallback timezone for floating ICS timestamps and the local dummy block calculation. Use an IANA zone such as `Europe/Budapest`, `UTC`, or `America/New_York`. |
| **`event`** | Object | No | | Controls how calendar events are mapped to programme title and description. See [ICS Event Mapping](#ics-event-mapping-event). |
| **`dummy`** | Object | No | | Controls optional gap filling with generated dummy programme entries. See [ICS Dummy Gap Filling](#ics-dummy-gap-filling-dummy). |
| **`include_cancelled`** | Bool | No | `false` | If set to `true`, calendar events with `STATUS:CANCELLED` are imported. By default, cancelled events are skipped. |
| **`max_events`** | Int | No | `50000` | Safety budget for encountered `VEVENT` blocks, including invalid or skipped events. Parsing stops when the budget is exhausted. Hard cap: `200000`. |
| **`max_download_bytes`** | Int | No | `10485760` | Maximum downloaded ICS size in bytes. The default is 10 MiB. Hard cap: `52428800` (50 MiB). |
| **`max_decompressed_bytes`** | Int | No | `20971520` | Maximum decompressed ICS size in bytes. The default is 20 MiB. Hard cap: `104857600` (100 MiB). |
`timezone` must be a valid IANA timezone known to Tuliprox, for example `UTC`, `Europe/Budapest`, or
`America/New_York`.
#### ICS Event Mapping (`event`)
The `event` block controls how ICS `VEVENT` properties are converted into EPG programme fields.
| **`enabled`** | Bool | No | `false` | Enables generation of dummy programme entries for gaps without real events. |
| **`title`** | String | No | `No programme entry` | Programme title for generated dummy entries. |
| **`description`** | String | No | empty | Programme description for generated dummy entries. |
| **`days_past`** | Int | No | `1` | Number of days before the current day for which dummy entries are generated. |
| **`days_future`** | Int | No | `14` | Number of days after the current day for which dummy entries are generated. |
| **`block_hours`** | Int | No | `4` | Size of the local dummy blocks in hours. The value must divide 24 evenly. For example: `1`, `2`, `3`, `4`, `6`, `8`, or `12`. |
| **`min_gap_minutes`** | Int | No | `1` | Minimum gap length required to generate a dummy entry. Smaller gaps are ignored to avoid tiny placeholder programmes caused by second-level timestamp differences. |
Dummy entries never replace real programme entries. They only fill uncovered time ranges.
`days_past` is capped at `30`; `days_future` is capped at `366`. Dummy titles use the same length cap as event summaries
and dummy descriptions use the same length cap as event descriptions.
Example with one real event from `03:30` to `05:00`:
```text
00:00 - 03:30 No Formula 1 session
03:30 - 05:00 Real calendar event
05:00 - 08:00 No Formula 1 session
08:00 - 12:00 No Formula 1 session
...
```
Example with an event aligned to a block boundary:
```text
00:00 - 04:00 No Formula 1 session
04:00 - 06:00 Real calendar event
06:00 - 08:00 No Formula 1 session
08:00 - 12:00 No Formula 1 session
...
```
If two events are adjacent, no dummy programme is inserted between them:
```text
10:00 - 11:00 Real calendar event
11:00 - 12:00 Real calendar event
12:00 - 16:00 No Formula 1 session
```
Dummy block boundaries are calculated in `ics.timezone`. Around daylight saving time changes, the local block labels
remain stable, for example `00:00 - 04:00`, `04:00 - 08:00`, and so on, while the corresponding UTC duration may differ.
If a local boundary does not exist during a spring-forward transition, it maps to the first real instant after the gap.
If a boundary is ambiguous during a fall-back transition, the earlier occurrence is used. Any resulting zero-length or
negative UTC interval is omitted, so generated dummy programmes remain ordered, non-overlapping, and continuous over
the real local day.
#### ICS Download and Security Notes
ICS sources use the same download infrastructure as playlist and XMLTV EPG downloads. This means that input-level
| `id` | Int | No | | Internal/generated alias ID. Normally omit this; Tuliprox assigns IDs from the input/alias order during config preparation. |
| `name` | String | Yes | | Unique alias name. Used for stable playlist UUID generation and consistent channel numbering across updates. |
| `url` | String | Yes | | Provider base URL, playlist URL, or `provider://<name>` reference. For M3U aliases, credentials can also be extracted from the URL query parameters. |
| `username` | String | Xtream | | Account username. Mandatory for regular Xtream YAML aliases. Optional for M3U aliases when credentials are embedded in the playlist URL. |
| `password` | String | Xtream | | Account password. Mandatory for regular Xtream YAML aliases. Optional for M3U aliases when credentials are embedded in the playlist URL. |
| `priority` | Int | No | `0` | Connection selection priority for this alias. Lower numbers have higher priority; negative numbers are allowed. |
| `max_connections` | Int | No | `0` | Allowed concurrent streams for this alias. In YAML, `0` means unlimited/no explicit limit. |
| `exp_date` | Mixed | No | | Account expiration. Supports `"YYYY-MM-DD HH:MM:SS"` interpreted as UTC or Unix timestamps in seconds. Used for status tracking and Panel API sync/renewal logic. |
| `enabled` | Bool | No | `true` | If `false`, this alias is ignored when Tuliprox builds the usable connection pool. |
| **`url`** | Provider base URL or full M3U playlist URL. Required in CSV rows. |
| **`name`** | **Crucial:** The first alias is automatically renamed with the `name` from the input definition (e.g., `my_provider_1` gets `my_provider`). This is necessary for stable playlist UUID generation and consistent channel numbering across updates. |
| **`username`** | Xtream or Stalker account username. For M3U CSV rows, Tuliprox can also extract credentials from the URL query parameters. |
| **`password`** | Xtream or Stalker account password. For M3U CSV rows, Tuliprox can also extract credentials from the URL query parameters. |
| **`mac_address`** | Stalker alias MAC address. Required by `mac_only` and `mac_plus_credentials`; optional for credentials-only authentication. |
| **`auth_mode`** | Stalker alias authentication mode: `auto`, `mac_only`, `credentials_only`, or `mac_plus_credentials`. |
| **`mag_preset`** | Stalker alias MAG profile: `generic_safe`, `mag250_legacy`, `mag254_strict`, or `ministra_modern`. |
| **`max_connections`** | Defines allowed concurrent streams. Default in CSV is **1**. |
| **`priority`** | Lower numbers = higher priority. `0` is higher than `1`. Negative numbers (e.g., `-1`) are allowed for top-tier priority. Items with the lowest values are processed first. |
| **`exp_date`** | Account expiration. Supports `"YYYY-MM-DD HH:MM:SS"` interpreted as UTC or Unix timestamps in seconds. Used for auto-cleanup or Panel API sync. |
| **`enabled`** | Enables/disables a CSV alias row. Empty values, `1`, `t`, and `true` are treated as enabled; `0`, `f`, or `false` disable the alias. |
| **`alias_pool`** | Object | | Controls the lifecycle of active aliases. |
| ↳ `size.min` | Mixed | `1` | Min accounts to keep. `number` or `auto`. If `auto`, it uses the count of enabled Tuliprox users (Active/Trial, not expired) mapped to this input's targets. |
| ↳ `size.max` | Mixed | `1` | Upper bound for aliases. If `auto`, checks are triggered upon user add/update. |
| ↳ `remove_expired` | Bool | `false` | If `true`, removes expired accounts from `source.yml` or batch CSVs during boot/update. (The root input is never removed). |
| `enabled` | Bool | No | `true` | If set to `false`, Tuliprox skips building this target during normal processing. This reduces CPU, disk, and upstream workload, but the target can still be selected explicitly via CLI target execution if matched by `-t`. |
| `name` | String | No | `default` | Logical target name. If not `default`, it must be unique. Unique names are important for selective execution (`-t <target_name>`) and for clearly separating output identities in Tuliprox's processing pipeline. |
| `processing_order` | Enum | No | `frm` | Defines execution order for **F**ilter, **R**ename, and **M**ap. This directly changes which intermediate state downstream steps operate on and can therefore materially alter the final playlist result. |
| `filter` | String | Yes | | Global filter DSL expression for the target. This determines which entries survive into the final target after the selected processing order has been applied. |
| `rename` | List | No | | Regex-based transformations applied to selected fields. This is commonly used to normalize channel/group labels before sorting, mapping, or export. |
| `mapping` | List | No | | References mapping IDs from `mapping.yml` for advanced transformation logic. This is where deep structural rewriting and metadata normalization can be applied. |
| `sort` | Object | No | | Defines ordering for groups and channels after transformations. This affects the final playlist structure seen by clients and can significantly improve navigation quality in IPTV players. |
| `options` | Object | No | | Target-level behavior switches such as logo suppression, duplicate removal, and shared live-stream handling. These options influence memory usage, playlist cleanliness, and reverse-proxy behavior. |
| `output` | List | Yes | | Mandatory list of output formats. A single target can generate multiple output representations (e.g., `xtream`, `m3u`, `strm`, `hdhomerun`) from the same transformed result set. |
| `favourites` | List | No | | Duplicates final transformed channels into dedicated favorite groups after processing is complete. This adds curated views without changing the original group structure. |
| `watch` | List | No | | Defines watched group patterns. If matching groups change during updates, Tuliprox emits Messaging events so operational changes become observable automatically. |
| `use_memory_cache` | Bool | No | `false` | If enabled, the final compiled playlist is cached in RAM. This reduces disk access and improves delivery speed, especially for M3U downloads, but increases memory consumption. |
---
### 3.2.1 `processing_order`
The processing order defines how Tuliprox applies:
* **F**ilter
* **R**ename
* **M**ap
Valid values are:
*`frm` (default)
*`fmr`
*`rfm`
*`rmf`
*`mfr`
*`mrf`
> **Note:** The selected processing order can change the final result significantly.
> For example, if renaming occurs before filtering, the filter must match the renamed state rather than the original
| `field` | Enum | Yes | | Field to transform, can be `group`, `title`, `name`, `caption` or `url`. This determines which part of the playlist entry Tuliprox rewrites before later stages such as sorting or final export. |
| `pattern` | String (Regex) | Yes | | Regular expression used to match the current value of the selected field. This enables structural normalization of inconsistent source naming schemes. |
| `new_name` | String | Yes | | Replacement string. It can reference regex capture groups via `$1`, `$2`, and so on. This allows Tuliprox to preserve selected original content while reformatting labels. |
#### Rename Example
Example:
```yaml
rename:
- field:group
pattern:'^DE(.*)'
new_name:'1. DE$1'
```
In above example, every group beginning with `DE` is renamed to start with `1.`, for example:
*`DE Sports` → `1. DE Sports`
*`DE Movies` → `1. DE Movies`
This can be useful for players that ignore provider order and perform their own alphabetical sorting.
> **Note:** The effective value that `rename` sees depends on `processing_order`.
> If mapping runs before renaming, your rename pattern must match the already mapped value rather than the original
> source value.
---
### 3.2.4 `mapping`
The `mapping` block references a list of mapping identifiers (IDs) defined in your [mapping files](./mapping-dsl.md) (
| `mapping` | List of Strings | No | | Ordered list of mapping IDs to apply. Each referenced mapping can perform deep transformations on the playlist structure, metadata, grouping, or labels, making this one of the most powerful target-level processing stages in Tuliprox. |
To define a new mapping IDs see details in chapter [Mapper DSL & Logic](./mapping-dsl.md).
---
### 3.2.5 `sort`
The `sort` block defines ordering rules for groups and channels.
| `match_as_ascii` | Bool | No | `false` | If enabled, Tuliprox normalizes accented characters during sorting comparisons. This improves deterministic ordering across multilingual playlists without modifying the original visible channel names. |
| `rules` | List | Yes | | Ordered list of sort rules. Each rule is evaluated against the playlist after transformation, and directly shapes the browsing order clients see in the final target. |
| `target` | Enum | Yes | | Defines whether the rule sorts `group` or `channel` entries. This changes whether Tuliprox reorders category containers or items within those categories. |
| `field` | String | Yes | | Sort field. For `channel`: `title`, `name`, `caption`, or `url`. For `group`: `group`. This determines which final-state value Tuliprox uses for ordering. |
| `filter` | String | Yes | | Filter expression defining which entries the rule applies to. This makes it possible to sort only selected subsets of the playlist instead of the entire target uniformly. |
| `order` | Enum | Yes | | `asc`, `desc`, or `none`. `none` preserves source order for matched entries and is useful when provider order should remain untouched. |
| `sequence` | List | No | | Ordered regex list used for index-based sorting. When present, Tuliprox prioritizes regex sequence position over `order`, enabling explicit semantic ordering such as quality tiers or curated group precedence. |
> **Note:** Sort rules must be written with the configured `processing_order` in mind,
> because sorting operates on the transformed state that exists at that point in the pipeline.
#### Sort Example
```yaml
sort:
match_as_ascii:false
rules:
- target:group
order:asc
filter:'Group ~ ".*"'
field:group
sequence:
- '^Freetv'
- '^Shopping'
- '^Entertainment'
- '^Sunrise'
- target:channel
order:asc
filter:'Group ~ ".*"'
field:title
sequence:
- '(?P<c1>.*?)\bUHD\b'
- '(?P<c1>.*?)\bFHD\b'
- '(?P<c1>.*?)\bHD\b'
- '(?P<c1>.*?)\bSD\b'
```
**Named Capture Groups** in `sequence`
To sort by specific parts of a value, use named capture groups such as:
| `ignore_logo` | Bool | No | `false` | Ignores `tvg-logo` and `tvg-logo-small` attributes. This reduces downstream device-side logo caching and can keep generated M3U playlists leaner for clients with limited storage or poor cache invalidation behavior. |
| `share_live_streams.hls` | Bool | No | `false` | Enables HLS live sharing for the new HLS cache proxy path. This is a configuration switch for the HLS cache feature and is independent from MPEG-TS stream sharing. |
| `share_live_streams.mpeg_ts` | Bool | No | `false` | Allows Tuliprox to share MPEG-TS live stream connections in reverse proxy mode. This can reduce upstream provider connection usage when multiple clients watch the same channel, but it increases memory usage per shared channel. |
| `remove_duplicates` | Bool | No | `false` | Attempts to remove duplicate entries by `url`. This improves playlist cleanliness and reduces confusing duplicates in the client-facing output. |
| `epg_output.lowercase_ids` | Bool | No | `false` | Canonicalizes visible technical EPG IDs with ASCII lowercase across M3U `tvg-id`, Xtream `epg_channel_id`, XMLTV channel/programme references, and EPG API responses. Changing this option requires a full target refresh. |
| `epg_output.lowercase_xmltv_display_names` | Bool | No | `false` | Applies Unicode lowercase exclusively to XMLTV `<display-name>` values during serialization. Playlist names, Xtream names, programme titles, and programme descriptions remain unchanged; persisted target data does not require rebuilding. |
| `force_redirect` | Bool | No | `false` | Optional redirect-related behavior switch. This influences how Tuliprox serves final stream delivery where redirect-style output handling is required by the deployment model. |
| `type` | Enum | Yes | | Output format type. Supported values include `xtream`, `m3u`, `strm`, and `hdhomerun`. This determines how Tuliprox serializes and serves the final playlist to downstream consumers. |
| `filter` | String | No | | Optional output-level filter applied after all target transformations. This allows Tuliprox to derive specialized output subsets from the same target without duplicating upstream processing logic. |
**Specific Output Properties** are defined for each type:
| `type` | Enum | Yes | | Must be `xtream`. Generates an Xtream-compatible API output backed by Tuliprox's processed data model. |
| `skip_live_direct_source` | Bool | No | `true` | If `true`, Tuliprox ignores provider `direct_source` values for live content. This keeps playback under Tuliprox's delivery logic and avoids client behavior differences caused by bypass URLs. |
| `skip_video_direct_source` | Bool | No | `true` | If `true`, Tuliprox ignores provider `direct_source` values for movies/VOD. This improves consistency across clients that otherwise may bypass Tuliprox for video playback. |
| `skip_series_direct_source` | Bool | No | `true` | If `true`, Tuliprox ignores provider `direct_source` values for series entries. This ensures Tuliprox stays in control of series playback URL generation and proxy behavior. |
| `update_strategy` | Enum | No | `instant` | `instant` writes changes immediately, while `bundled` batches write operations. This directly trades off freshness versus disk I/O load during background metadata enrichment and output maintenance. |
| `trakt` | Object | No | | Trakt.tv integration block. Tuliprox can fetch Trakt lists, fuzzy-match them against playlist entries, and inject matched VOD or series entries into generated virtual categories. |
| `filter` | String | No | | Optional output-level filter for the Xtream export only. Useful when the same target should expose different subsets to different output formats. |
> **Note:** IPTV players vary in how they resolve streams: some use the direct-source attribute, while others
> reconstruct URLs
> from server metadata. To ensure Tuliprox maintains control over the stream routing (Proxy/Redirect),
> the Direct Source Handling (skip_*_direct_source) attributes default to true.
>
> **⚠️ Warning:** Setting `skip_*_direct_source` to `false` forces the player to use the provider's original
`direct-source` URL.
> This effectively **bypasses Tuliprox**, which will disable internal features like connection tracking,
> IP masking, and failover logic for those streams.
#### `trakt` Object in Xtream Output
Trakt.tv is an online platform for tracking, organizing, and discovering movies and TV shows.
Tuliprox can query Trakt lists and match playlist entries using Jaro-Winkler-style fuzzy matching.
Matching entries are then added to new virtual categories inside the Xtream output.
| `api.api_key` | String | Yes | | Trakt API key used for authenticated access. Without a valid key, Tuliprox cannot fetch remote list content. |
| `api.version` | String | No | `"2"` | API version header value. This ensures Tuliprox formats requests against the correct Trakt API version. |
| `api.url` | String | No | `https://api.trakt.tv` | Base API URL for Trakt requests. This defines the remote endpoint Tuliprox queries for list data. |
| `api.user_agent` | String | No | | Optional `User-Agent` used for Trakt API requests. This can help satisfy API gateway expectations or deployment-specific request policies. |
| `lists[].user` | String | Yes | | Trakt username owning the list. This identifies which account namespace Tuliprox fetches list data from. |
| `lists[].list_slug` | String | Yes | | Trakt list slug. Combined with `user`, this uniquely identifies the remote list to load. |
| `lists[].category_name` | String | Yes | | Name of the generated virtual category inside Tuliprox's Xtream output. This controls where matched entries appear to clients. |
| `lists[].content_type` | Enum | Yes | | `vod` or `series`. This determines which class of playlist entries Tuliprox will attempt to match and inject into the generated category. |
| `lists[].tmdb_only` | Bool | No | `false` | If `true`, only exact TMDB-id matches are accepted for this list, disabling title/year fuzzy fallback and reducing false positives. |
| `lists[].fuzzy_match_threshold` | Integer | No | | Fuzzy matching threshold for title matching. Higher values reduce false positives but may miss loosely matching items. |
| `charts[]` | List | No | `[]` | Public Trakt chart definitions. Unlike `lists[]`, these are system charts and do not have a user/list owner. |
| `charts[].kind` | Enum | Yes | | `movies` or `shows`. Aliases such as `movie`, `vod`, `show`, `series`, and `tvshows` are accepted. |
| `charts[].chart` | Enum | Yes | | Public chart to fetch. MVP supports `trending` and `popular`. |
| `charts[].category_name` | String | Yes | | Name of the generated virtual category inside Tuliprox's Xtream output. |
| `charts[].tmdb_only` | Bool | No | `false` | If `true`, only exact TMDB-id matches are accepted. This is recommended for dynamic charts to avoid fuzzy false positives. |
| `charts[].fuzzy_match_threshold` | Integer | No | | Fuzzy matching threshold for chart title matching when `tmdb_only` is not enabled. |
The `charts[]` MVP intentionally supports only public, non-OAuth Trakt charts. User-specific recommendations and
account-scoped history feeds are not fetched by this block.
| `type` | Enum | Yes | | Must be `m3u`. Generates a traditional playlist file suitable for IPTV players and related clients. |
| `filename` | String | No | | Optional custom output filename. This affects how Tuliprox writes or exposes the generated playlist artifact. |
| `include_type_in_url` | Bool | No | `false` | If enabled, Tuliprox adds the stream type (`live`, `movie`, `series`) into generated stream URLs. This can improve downstream routing clarity and compatibility with clients that distinguish path structure by media type. |
| `mask_redirect_url` | Bool | No | `false` | If enabled, Tuliprox uses URLs from `api-proxy.yml` for users operating in `redirect` proxy mode. This is important for multi-provider failover or cycling setups where exposing the provider URL directly would bypass Tuliprox's routing logic too early. |
| `filter` | String | No | | Optional M3U-only post-transformation filter. This allows M3U consumers to receive a narrower subset than other output formats derived from the same target. |
> **Note:** `mask_redirect_url` should be enabled if you use multiple providers and want Tuliprox to preserve
> redirect-mode
> routing and cycling behavior without exposing the direct upstream endpoint in the initial playlist URL.
| `type` | Enum | Yes | | Must be `strm`. Generates filesystem-based `.strm` references instead of a network playlist format. |
| `directory` | String | Yes | | Target directory where `.strm` files are written. This is the root Tuliprox manages for exported media stubs and must be chosen carefully to avoid overlap with real media directories. |
| `username` | String | No | | Optional username context used when generating stream references. This affects which user-specific URL or access context Tuliprox embeds into the exported `.strm` files. |
| `underscore_whitespace` | Bool | No | `false` | Replaces whitespace with `_` in paths and filenames. This improves compatibility with environments or scrapers that prefer filesystem-safe, normalized naming. |
| `cleanup` | Bool | No | `false` | If enabled, Tuliprox removes orphaned output files from the STRM directory. This keeps the export directory synchronized with the target, but can delete files if the directory points to an existing media folder. |
| `flat` | Bool | No | `false` | If enabled, Tuliprox creates a flatter directory structure. This changes how categories and group information are represented on disk and can simplify some media-server imports. |
| `strm_props` | List | No | | Stream property lines inserted into `.strm` files, mainly for Kodi player behavior. This allows low-level playback hints to be embedded directly into generated files. |
| `add_quality_to_filename` | Bool | No | `false` | Appends detected media quality tags such as `[1080p 4K HEVC HDR]` to the filename. This improves visibility in library UIs but depends on prior probing/enrichment data being available. |
| `filter` | String | No | | Optional STRM-only output filter. Useful when only a subset of the target should be materialized as filesystem entries. |
| `type` | Enum | Yes | | Must be `hdhomerun`. Exposes the target through Tuliprox's HDHomeRun emulation layer for tuner-style discovery by clients such as Plex or Jellyfin. |
| `device` | String | Yes | | Must match a device name defined in `config.yml`. This links the playlist target to a specific emulated tuner endpoint. |
| `username` | String | Yes | | Must match a user from `api-proxy.yml`. This determines which account context, access restrictions, and connection limits apply when clients consume the lineup through the tuner interface. |
| `use_output` | Enum | No | | Selects whether the HDHomeRun stream URLs are based on `m3u` or `xtream` output behavior. This affects how playback URLs are generated and which delivery semantics back the tuner lineup. |
| `cluster` | String | No | | Optional logical cluster, for example `series`. This influences how Tuliprox groups the duplicated entries internally for output generation. |
| `group` | String | Yes | | Name of the favorite group created in the final playlist. This adds a curated access path without removing the original group membership. |
| `filter` | String | Yes | | Filter expression selecting which final entries should be duplicated into the favorites group. This operates on the transformed end state rather than the original raw input. |
| `match_as_ascii` | Bool | No | `false` | If enabled, Tuliprox normalizes accented characters during matching. This improves filter matching robustness across multilingual names while preserving the original visible title in output. |
---
### 3.2.9 Watch (`watch`)
For each target with a *unique name*, you can define watched groups.
It is a list of group patterns Tuliprox monitors for content changes during updates.
If matching groups gain or lose channels, Tuliprox emits a Messaging event such as:
| `group` | String (Regex) | Yes | | Regex pattern matched against final group names. This allows Tuliprox to detect meaningful content changes in selected areas of the playlist and notify operators automatically through the configured messaging backends. |
