Readwise Reader

https://docs.readwise.io/reader
Admin
Readwise Reader is an all-in-one reading app that unifies articles, newsletters, EPUBs, PDFs,...
Tokens:48,860
Snippets:517
Trust Score:9.9
Update:7 months ago
Show doc for...
Tokens:
Raw
### Crafting Custom Summary Prompts with Specific Tones

Source: https://docs.readwise.io/reader/guides/ghostreader/tips-tricks

This example shows how to create a Ghostreader prompt that summarizes text in a specific style, inspired by fanfiction summaries. It guides the AI to adopt a playful, honest, and self-aware tone, focusing on enticing elements rather than a strict plot outline, and using fandom-inspired tags. This method helps in achieving desired tones for summaries.

```markdown
On a new line, in 1-3 concise sentences, summarize the story in the style of a fanfiction summary. Aim for a playful, honest, and self-aware tone, as if talking to a fellow fan. Instead of a strict plot outline, focus on enticing elements like relationships, conflicts, and unique quirks, hinting at major themes or twists without giving too much away. Use fandom-inspired tags to highlight key dynamics or tropes readers might be looking for—think “Enemies to Lovers,” “Found Family,” or “Angst with a Happy Ending”. Emphasize emotion, humor, or drama, leaning into hyperbole where fitting without verging into melodrama or "cringe".

```

--------------------------------

### Filter by Minutes (Reading Time)

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

Filters documents based on their estimated reading time in minutes. This allows for selection of content within a specific time range.

```text
minutes:15
```

--------------------------------

### Filter by Highlights

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

Filters documents based on the number of highlights made within them. Use `has:highlights` to see all documents with at least one highlight.

```text
highlights:5
```

```text
has:highlights
```

--------------------------------

### Prompting with Document Summary - Jinja2

Source: https://docs.readwise.io/reader/guides/ghostreader/reference

An example prompt that incorporates the document's summary using the `document.summary` Jinja2 variable. This allows the AI to consider the existing summary when generating a response.

```jinja2
I'm about to read a document entitled "{{ document.title }}" written by {{ document.author }} having the following summary:
===
{{ document.summary }}
===


```

--------------------------------

### Classify Content with Options

Source: https://docs.readwise.io/reader/guides/ghostreader/reference

This example shows how to use the 'classify' subroutine to detect sentiment, labels, or tags within text. It demonstrates options like 'classes', 'hypothesis', 'multi_label', and 'threshold', and how to join the results.

```liquid
{{ document.content | truncate | classify(classes=["news", "sports"], hypothesis="This news article was published under {}", multi_label=True) | join: ', ' }}

```

--------------------------------

### Get User Input with Input

Source: https://docs.readwise.io/reader/guides/ghostreader/reference

The `input` utility allows prompts to accept user-provided text. The text within the parentheses serves as a placeholder prompt for the user. Whitespace can be stripped using the minus symbol.

```jinja
{%- set query = input("Instructional placeholder text goes here") -%}
```

--------------------------------

### Extract User Selection and Context

Source: https://docs.readwise.io/reader/guides/ghostreader/reference

This example shows how to display text selected by the user, along with the sentence and paragraph containing that selection. It uses the 'selection' variable and its properties 'sentence' and 'paragraph'.

```liquid
I just came across the word or phrase "{{ selection }}" as used in the following sentence: "{{ selection.sentence }}"

Within the document, I selected this text:
{{ selection }}

That selection is contained in the paragraph:
{{ selection.paragraph }}

```

--------------------------------

### Prompting with Document Title and Author - Jinja2

Source: https://docs.readwise.io/reader/guides/ghostreader/reference

An example of a prompt that uses Jinja2 variables `document.title` and `document.author` to personalize the request. It asks the AI to generate questions based on the document's title and author.

```jinja2
I'm about to read a document entitled "{{ document.title }}" written by {{ document.author }}. I want you to identify the 3 most interesting questions I should consider while reading.

```

--------------------------------

### Automatically Tag Ghostreader Responses with Action Tags

Source: https://docs.readwise.io/reader/guides/ghostreader/tips-tricks

This example demonstrates how to automatically apply highlight tags to Ghostreader responses by prefixing the tag name with a dot (e.g., '.trivia'). This is useful for consistent tagging when syncing highlights to Readwise. The tag is not visible in Reader but is applied upon syncing.

```markdown
Please write a brief encyclopedia entry for "{{ selection }}". For context, I encountered this term in the following sentence: {{ selection.sentence }}

Here's an example of the length and tone to use for your brief encyclopedia entry:
===
.trivia
Unix philosophy: The philosophy that small, modular programs, which can be easily combined to perform complex tasks, are more effective and easier to maintain than large, monolithic programs.
===

IMPORTANT: Use the same formatting for the response as seen in the example. ALWAYS begin with ".trivia"

```

--------------------------------

### Conditional Logic with Document Domain - Jinja2

Source: https://docs.readwise.io/reader/guides/ghostreader/reference

Provides an example of using Jinja2 to check if a document originates from a specific domain, such as 'nytimes.com'. This enables conditional prompt logic based on the source website.

```jinja2
{% if "nytimes.com" in document.domain %}

{% endif %}
```

--------------------------------

### LLM Summarization Instructions

Source: https://docs.readwise.io/reader/guides/ghostreader/custom-prompts

These are plain language instructions provided to the LLM to guide the summarization process. They emphasize conciseness, readability, and capturing the main idea, aiming to keep the LLM focused and prevent it from going off-topic.

```Plain Text
IMPORTANT: Write no more than THREE sentences. Each sentence should be short and easy-to-read. Use words sparingly and please capture the big idea.
```

--------------------------------

### Jinja2 If-Elif-Else Statement for Document Titles

Source: https://docs.readwise.io/reader/guides/ghostreader/custom-prompts

This Jinja2 example demonstrates a more complex conditional logic using if, elif, and else statements to tailor prompt output based on the document title. It checks for specific phrases within the title to determine which section of the prompt to execute, allowing for nuanced responses.

```jinja
{% if "?" in document.title %}
...
{% elif "Twitter List" in document.title %}
...
{% else %}
...
{% endif %}
```

--------------------------------

### Displaying Document Language - Jinja2

Source: https://docs.readwise.io/reader/guides/ghostreader/reference

An example of using the `document.language` Jinja2 variable to display the language of the document being processed. This can be useful for context-aware prompts.

```jinja2
I'm reading a document in {{ document.language }}.

```

--------------------------------

### Emoji Bullet Point Summary (Jinja)

Source: https://docs.readwise.io/reader/guides/ghostreader/prompt-library

This prompt generates a concise, three-sentence summary of a document, with each sentence starting with a unique emoji. It's designed for easy readability, akin to explaining concepts to an 11-year-old. Jinja templating is used to dynamically include document details and content, with a condition to process longer documents using central sentences.

```jinja
Write three short sentences briefing me with the insights I need to know for the following document. Each sentence should start with a unique emoji and ending in a newline:

"""
Title: {{ document.title }}
Author: {{ document.author }}
Domain: {{ document.domain}}
{% if (document.content | count_tokens) > 2000 %}
{{ document.content | central_sentences | join('\n\n') }}
{% else %}
{{ document.content }}
{% endif %}
"""

IMPORTANT: Write no more than THREE sentences. Each sentence should be short, matter of fact, and easy-to-read, like talking to an 11 year old. I can't read the full text so I rely on your brief to pretend as if I had read it. Start each sentence with a unique emoji, and end it with a newline.
```

--------------------------------

### Filter by Not Operator

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

This operator filters results where the parameter does not match the specified input. It can be used with any data type to exclude specific items.

```text
tags__not:"fiction"
```

--------------------------------

### Filter by Contains Operator

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

This operator filters text-based parameters to find results where the content includes the specified text anywhere within it. It is case-sensitive.

```text
title__contains:"Artificial Intelligence"
```

--------------------------------

### Filter by Less Than Operator

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

This operator filters results where a numerical parameter is strictly less than the specified value. It is useful for setting maximum thresholds.

```text
minutes__lt:10
```

--------------------------------

### Filter by Saved Count

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

Filters documents based on how many times they have been saved to your account. This is useful for identifying popular or frequently accessed content.

```text
saved_count:100
```

--------------------------------

### Jinja2 Prompt for Document Summarization

Source: https://docs.readwise.io/reader/guides/ghostreader/custom-prompts

This Jinja2 template instructs an AI model to summarize a document into three concise sentences. It dynamically includes document metadata (title, author, domain) and content, with logic to handle varying document lengths to optimize for AI prompt limits. It also includes specific formatting instructions for the output.

```jinja2
{#- BACKGROUND: This prompt instructs ChatGPT to summarize the document into three information-dense sentences. It's intended to be used after you're done reading. If you want your summaries in a language other than English, we recommend rewriting the entire prompt in the target language. You can get quite creative using combinations of logic and variables to enable all kinds of creative use cases. See the documentation for details and examples. -#}
Write three easy-to-read sentences summarizing the following text:

===
Title: {{ document.title }}
Author: {{ document.author }}
Domain: {{ document.domain}}
{#- The if-else logic below checks if the document is very long, long, or short in order to not exceed the GPT "prompt window". We highly recommend not changing this unless you know what you're doing. -#}
{% if (document.content | num_tokens) > 25000 %}
{{ document.html | central_paragraphs | join('\n\n') }}
{% elif (document.content | num_tokens) > 2500 %}
{{ document.content | central_sentences | join('\n\n') }}
{% else %}
{{ document.content }}
{% endif %}

IMPORTANT: Write no more than THREE sentences. Each sentence should be short and easy-to-read. Use words sparingly and please capture the big idea.

```

--------------------------------

### Filter by Annotation Presence

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

Filters documents based on the presence of specific annotations. Accepts 'highlights', 'tags', or 'notes' to find documents with those annotations.

```text
has:tags
```

```text
has:notes
```

--------------------------------

### Filter by Exact Match Operator

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

This operator filters results where the parameter content exactly matches the specified text. It is case-sensitive and requires an exact match.

```text
author__exact:"Isaac Asimov"
```

--------------------------------

### Filter by Before Date Operator

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

This operator filters results where a date parameter is earlier than the specified date or time frame. It accepts various date formats.

```text
published__before:"2023-01-01"
```

--------------------------------

### Generate Read Time Justification Prompt (Jinja)

Source: https://docs.readwise.io/reader/guides/ghostreader/prompt-library

This Jinja template prompt instructs the AI to generate a three-sentence summary justifying why a user should read a given document. It handles conditional logic for document content length and uses filters to extract central sentences if the content is extensive. The output is designed to be concise and capture the main idea.

```jinja
Write three easy-to-read sentences justifying why I should read the following text:
"""
Title: {{ document.title }}
Author: {{ document.author }}
Domain: {{ document.domain}}
{% if (document.content | count_tokens) > 2000 %}
{{ document.content | central_sentences | join('\n\n') }}
{% else %}
{{ document.content }}
{% endif %}
"""
IMPORTANT: Write no more than THREE sentences. Each sentence should be short and easy-to-read. Use words sparingly and please capture the big idea worth reading.
```

--------------------------------

### Filter by After Date Operator

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

This operator filters results where a date parameter is later than the specified date or time frame. It accepts various date formats.

```text
saved_at__after:"2024-03-15T10:00:00Z"
```

--------------------------------

### Apply Multiple Tags Based on Domain and Title (Ghostreader)

Source: https://docs.readwise.io/reader/guides/ghostreader/tips-tricks

This example expands on conditional tagging by incorporating multiple `elif` conditions. It tags New York Times articles with 'news' and YouTube videos related to AI (based on title keywords and category) with 'ai'. The `{# noghost #}` comment ensures that the LLM only processes relevant documents, preventing unwanted tags.

```ghostreader
{% if "nytimes.com" in document.domain %}
{# noghost #}
news
{% elif ("chatgpt" or "openai" in document.title) and "video" in document.category %}
{# noghost #}
ai
{% endif %}

```

--------------------------------

### Filter by Less Than or Equal To Operator

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

This operator filters results where a numerical parameter is less than or equal to the specified value. It is useful for setting inclusive maximum thresholds.

```text
highlights__lte:3
```

--------------------------------

### Filter by Greater Than or Equal To Operator

Source: https://docs.readwise.io/reader/guides/filtering/syntax-guide

This operator filters results where a numerical parameter is greater than or equal to the specified value. It is useful for setting inclusive minimum thresholds.

```text
saved_count__gte:20
```

--------------------------------

### Generate Highlight-Based Summary

Source: https://docs.readwise.io/reader/guides/ghostreader/prompt-library

This prompt generates a four-sentence summary of key takeaways from a document, based on user highlights. It includes logic to handle documents with more than 2000 tokens and ensures the prompt only runs on documents with existing highlights.

```jinja
{% if (document.content | count_tokens) > 2000 %}
{{ document.content | central_sentences | join('\n\n') }}
{% else %}
{{ document.content }}
{% endif %}

The most important pieces from this document are:

{% for highlight in document.highlights %}
{{ highlight.content }} {{ "\n\n" }}
{% endfor %}

Please write four easy-to-read sentences that capture the key takeaways of this article, based on what I've already highlighted. The important pieces represent the key takeaways of the document. Please begin by saying, "Based on your highlights, here are your key takeaways from this document:" Then share THREE key takeaways.

IMPORTANT:  Be sure to add paragraph breaks between each section.
```

--------------------------------

### Continue Reading Filtered View Query

Source: https://docs.readwise.io/reader/guides/filtering/default-views

This query defines the 'Continue reading' filtered view. It shows items started in the past week, not yet archived, and with more than 5% progress.

```text
progress__gt:5 AND last_opened__after:"1 week ago" AND (in:inbox OR in:later)
```
Tokens:
Raw