docs: remove AI-sounding style (em dashes, emojis, marketing tone, bare URLs) - #502

🏁 Script executed:

printf '%s\n' '--- applicable repository guidance ---'
find /tmp/coderabbit-repo-knowledge/deepnote-deepnote-4f22e1a3 -maxdepth 2 -type f -name '*.md' -print
printf '%s\n' '--- target document ---'
cat -n docs/conda-environment.md

Repository: deepnote/deepnote

Length of output: 6444

🌐 Web query:

Official Conda documentation: distinguish Miniconda installer from conda channels and package sources

💡 Result:

In the context of the conda ecosystem, the terms "Miniconda," "conda channels," and "package sources" refer to distinct concepts related to how software is distributed, acquired, and managed [1][2]. Miniconda Installer Miniconda is a minimal software installer that provides the conda package and environment manager [3][4]. It includes only a base set of components (typically conda, Python, and essential dependencies), providing a lightweight, "clean slate" foundation [1][5]. Users install Miniconda to get the command-line tool needed to manage environments and packages, after which they must manually install additional software as needed [5][6]. It is distinct from the full Anaconda Distribution, which comes pre-installed with hundreds of data science packages [5]. Conda Channels Conda channels are the remote, web-accessible repositories where conda packages are stored and managed [7][8]. When you use the conda command to install a package, it searches these designated channels to locate, download, and install the required software files [1][8]. Channels act as the distribution hubs for the ecosystem [2]. Package Sources The term "package sources" refers to the broader concept of where packages originate, which are implemented via channels [1]. While channels serve as the specific locations (URLs) that the conda tool queries, they represent different types of package sources [2]: - Default Channels: Often maintained by specific organizations (e.g., the 'defaults' channel maintained by Anaconda, Inc.), which may be subject to specific terms of service or licensing [7][2]. - Community Channels: Repositories like 'conda-forge' or 'bioconda' that are maintained by community volunteers and provide open-source access to a vast array of packages [1][2]. - Private or Internal Channels: Custom locations that organizations or individuals can configure to host their own private packages [7][9]. In summary, Miniconda is the software used to perform installations, while channels are the specific remote repositories (package sources) from which that software fetches the packages [1][5]. Users can configure their local.condarc file to specify which channels (sources) conda should search and in what priority [10][8].

Citations:

• 1: https://conda.org/learn/faq/
• 2: https://conda.org/blog/2024-08-14-conda-ecosystem-explained
• 3: https://docs.conda.io/projects/conda/en/latest/index.html
• 4: https://www.anaconda.com/docs/getting-started/miniconda/install/overview.md
• 5: https://www.anaconda.com/docs/getting-started/concepts/anaconda-or-miniconda
• 6: https://docs.conda.io/docs/user-guide/install/
• 7: https://docs.conda.io/projects/conda/en/latest/user-guide/concepts/channels.html
• 8: https://www.anaconda.com/docs/getting-started/concepts/what-is-a-channel
• 9: https://docs.conda.io/projects/conda/en/25.5.x/user-guide/concepts/channels.html
• 10: https://docs.conda.io/projects/conda/en/stable/configuration.html

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

• docs/getting-started.md#L41-L44 (this comment)
• docs/index.md#L8-L8

🏁 Script executed:

printf '%s\n' '--- applicable repository guidance ---'
find /tmp/coderabbit-repo-knowledge/deepnote-deepnote-4f22e1a3 -type f -name '*.md' -print | sort
printf '%s\n' '--- target lines ---'
cat -n docs/security-overview.md | sed -n '45,65p'
printf '%s\n' '--- nearby provider references ---'
rg -n -i 'anthropic|openai|zero.?data|30 days|customer data|AWS environment|customer.s AWS' docs README.md .github 2>/dev/null | head -120

Repository: deepnote/deepnote

Length of output: 10928

🏁 Script executed:

printf '%s\n' '--- repository-wide documentation guidance ---'
cat /tmp/coderabbit-repo-knowledge/deepnote-deepnote-4f22e1a3/conventions/repo-wide.md
printf '%s\n' '--- documentation guidance ---'
cat /tmp/coderabbit-repo-knowledge/deepnote-deepnote-4f22e1a3/learnings/docs.md
printf '%s\n' '--- relevant repository learning ---'
cat /tmp/coderabbit-repo-knowledge/deepnote-deepnote-4f22e1a3/learnings/repo-wide.md
printf '%s\n' '--- AI privacy policy text ---'
cat -n docs/ai-data-privacy.md | sed -n '1,40p'

Repository: deepnote/deepnote

Length of output: 5959

🌐 Web query:

Anthropic zero data retention agreement applies products approved enterprise API use exceptions official

💡 Result:

Anthropic's Zero Data Retention (ZDR) is a contractual arrangement that prevents Anthropic from storing customer prompts or model responses at rest after an API response is returned, except as required to comply with law or combat misuse and harm (e.g., enforcing Usage Policies) [1][2][3]. ZDR is not a default setting and must be requested and approved on a per-organization basis through the Anthropic sales team [1][2][4]. The policy applies to the following products, subject to approval and specific enablement: 1. Eligible Anthropic APIs [2][3] 2. Anthropic products using a Commercial organization API key (including Claude Code accessed via the API) [2][3] 3. Claude Code when used through a Claude for Enterprise plan (with ZDR specifically enabled for that organization) [1][2][4] Crucially, standard Claude Teams and Claude Enterprise chat product interfaces are not eligible for ZDR [1][3]. For Claude for Enterprise, administrators have control over standard data retention settings (with a 30-day minimum) rather than ZDR [3][5]. Additionally, enablement of ZDR is not automatic for new organizations created under an existing account; it must be requested separately for each new organization [1][4]. Note that for ZDR-eligible arrangements, Anthropic still retains User Safety classifier results to enforce its Usage Policy [2]. Organizations using Claude via third-party cloud platforms (e.g., Amazon Bedrock, Google Cloud, or Microsoft Foundry) should refer to those specific platforms' policies, even if using an Anthropic model [1][4][6]. top_results: [1][2][4][3][5]

Citations:

• 1: https://platform.claude.com/docs/en/manage-claude/api-and-data-retention
• 2: https://privacy.claude.com/en/articles/8956058-i-have-a-zero-data-retention-agreement-with-anthropic-what-products-does-it-apply-to
• 3: https://www.getvoibe.com/resources/claude-api-data-retention/
• 4: https://code.claude.com/docs/en/zero-data-retention
• 5: https://code.claude.com/docs/en/data-usage
• 6: https://support.claude.com/en/articles/15425996-data-retention-practices-for-covered-models

🌐 Web query:

OpenAI API data usage policies 30 days retention endpoint exceptions zero data retention official

💡 Result:

OpenAI's standard data retention policy for API usage dictates that customer content—such as prompts and responses—is retained for up to 30 days in abuse monitoring logs to facilitate safety and abuse detection [1][2]. This retention period applies by default unless a longer period is legally required or necessary to protect services from harm [1]. Zero Data Retention (ZDR) is a control that, once approved and configured, excludes customer content from these abuse monitoring logs [1][3]. ZDR is subject to prior approval by OpenAI and the acceptance of additional requirements [1]. Key aspects of ZDR include: Endpoint Eligibility: ZDR is not available for all endpoints [1]. Stateless endpoints, such as /v1/chat/completions, /v1/responses, and /v1/realtime, are generally eligible for ZDR [1][4]. Conversely, stateful endpoints that support persistent sessions or long-term management—such as the /v1/conversations, /v1/conversations/items, and Assistants-related APIs—are generally ineligible for ZDR, and content processed through them may be retained until the customer deletes it [1][4][3]. Legal and Safety Exceptions: Even with ZDR enabled, OpenAI maintains mandatory exceptions for legal compliance [1][4]. Specifically, OpenAI is required by law to report potential instances of child sexual abuse material (CSAM), and data flagged for such risks will be retained regardless of ZDR status to support manual review and mandatory reporting [4]. Data for Training: For API customers, OpenAI does not use inputs or outputs to train its models by default [5]. This protection is standard for the API platform, regardless of whether ZDR is enabled [5]. Management: Organizations can manage their retention settings via the OpenAI API, where they may configure project-level retention types such as zero_data_retention or modified_abuse_monitoring [6]. Customers interested in ZDR should contact OpenAI sales to discuss eligibility and requirements [3].

Citations:

• 1: https://developers.openai.com/api/docs/guides/your-data
• 2: https://community.openai.com/t/data-retention-for-batches/770572/5
• 3: https://www.spotdev.co.uk/blog/openai-zero-data-retention
• 4: https://data-today.net/openai-zero-data-retention-frontier-models/
• 5: https://openai.com/policies/how-your-data-is-used-to-improve-model-performance/
• 6: https://developers.openai.com/api/reference/resources/admin/subresources/organization/subresources/projects/subresources/data_retention/methods/update

🏁 Script executed:

printf '%s\n' '--- repository guidance ---'
for f in /tmp/coderabbit-repo-knowledge/deepnote-deepnote-4f22e1a3/*/*.md; do
  printf '\n--- %s ---\n' "$f"
  head -80 "$f"
done
printf '%s\n' '--- target file ---'
cat -n docs/security-overview.md | sed -n '40,65p'

Repository: deepnote/deepnote

Length of output: 13586

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.