Gorse 05 Config Migration

What it does

When your agent uses it

• Gorse 0.5 recommendations are poor/empty despite having data
• Collaborative filtering appears disabled even though [recommend.collaborative] section exists
• Migrating Gorse config from 0.4 to 0.5
• Old config keys like [recommend.popular]

Example prompts

• “/gorse-05-config-migration”

Workflow steps

5 steps, taken from the first numbered list in SKILL.md.

1. 1[[recommend.non-personalized]] uses Expr language for score/filter functions
2. 2[[recommend.item-to-item]] types: "tags" (label similarity), "users" (collaborative), "embedding"
3. 3[[recommend.user-to-user]] types: "items" (shared items), "tags", "embedding"
4. 4[recommend.ranker] types: "none", "fm" (factorization machines), "llm"
5. 5Custom feedback types: Any string is valid in positive_feedback_types (e.g., "extended_view", "bookmark")

What it can do on your machine

Read from SKILL.md and the folder at commit 6487b05. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  No scripts in the folder and no shell commands in SKILL.md (its code samples are toml).

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  Links to these hosts (documentation or services it may open):

  • github.com
  • gorse.io

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Gorse 05 Config Migration loads about 1.5k tokens when it runs. Until then it costs about 152 tokens; SKILL.md has 351 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~152

When it runs · the whole SKILL.md, loaded when a task matches

~1.5k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from divinevideo/divine-mobile at commit 6487b05, republished under its MPL-2.0 licence (© divinevideo). 351 words, ~1,498 tokens.

name

gorse-05-config-migration

description

Fix Gorse 0.5.x recommendation engine silently disabled or degraded due to config incompatibility with 0.4. Use when: (1) Gorse 0.5 recommendations are poor/empty despite having data, (2) collaborative filtering appears disabled even though [recommend.collaborative] section exists, (3) migrating Gorse config from 0.4 to 0.5, (4) old config keys like [recommend.popular], [recommend.neighbors], [recommend.online], [recommend.offline] are being used with Gorse 0.5. Critical: type="mf" must be explicitly set in [recommend.collaborative] or CF is silently disabled (default "none").

author

Claude Code

version

1.0.0

date

2026-02-28

# 0.4 style (BROKEN in 0.5 — CF silently disabled):
[recommend.collaborative]
fit_period = "60m"
fit_epoch = 100

# 0.5 style (WORKING — must add type = "mf"):
[recommend.collaborative]
type = "mf"
fit_period = "60m"
fit_epoch = 100

[recommend]
cache_size = 100
cache_expire = "72h"
context_size = 100

[recommend.data_source]
positive_feedback_types = ["reaction", "comment", "repost"]
read_feedback_types = ["view"]
positive_feedback_ttl = 0
item_ttl = 0

# Collaborative filtering — MUST set type = "mf" to enable
[recommend.collaborative]
type = "mf"
fit_period = "60m"
fit_epoch = 100
optimize_period = "360m"
optimize_trials = 10

[recommend.collaborative.early_stopping]
patience = 10

# Trending/popular — TOML array syntax [[...]] for multiple leaderboards
[[recommend.non-personalized]]
name = "trending_weekly"
score = "count(feedback, .FeedbackType == 'reaction') + count(feedback, .FeedbackType == 'comment') * 2"
filter = "(now() - item.Timestamp).Hours() < 168"

# Content-based item similarity (requires items to have Labels)
[[recommend.item-to-item]]
name = "similar_content"
type = "tags"         # matches on item Labels

# Collaborative item similarity
[[recommend.item-to-item]]
name = "also_liked"
type = "users"        # users who liked X also liked Y

# User similarity
[[recommend.user-to-user]]
name = "similar_users"
type = "items"        # users who share liked items

# FM ranker blends all candidate sources
[recommend.ranker]
type = "fm"           # or "llm" for LLM-based reranking
recommenders = ["latest", "collaborative", "non-personalized/trending_weekly", "item-to-item/similar_content", "item-to-item/also_liked", "user-to-user/similar_users"]
fit_period = "60m"
fit_epoch = 100

[recommend.ranker.early_stopping]
patience = 10

[recommend.fallback]
recommenders = ["latest"]

[recommend.replacement]
enable_replacement = true
positive_replacement_decay = 0.8
read_replacement_decay = 0.6

© divinevideo, MPL-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .agents/skills/gorse-05-config-migration of divinevideo/divine-mobile.

Open the folder on GitHubat commit 6487b05