# Tier 1 - Inline Context and File Attachment

The simplest and most immediately accessible form of context management requires no technical configuration at all. It is available to every user of GLBNXT Workspace from the first conversation.

#### The Document Upload Button

Both OpenWebUI and LibreChat within GLBNXT Workspace provide a file attachment capability accessible via the plus sign adjacent to the message input field. Selecting "Add Document" allows you to upload a file directly into the current conversation. The contents of that file are processed and added to the context window for that session.

This single capability covers an enormous range of knowledge work tasks. Summarising a report, extracting key data points from a spreadsheet, reviewing a contract, answering questions about a policy document, comparing two versions of a proposal: all of these are within reach with nothing more than a file upload and a well-formed prompt.

The key constraints to understand are:

**Session scope.** A document uploaded in one conversation is not available in another. The attachment exists within the context of the current session only. If you close the conversation and open a new one, you will need to re-upload the document.

**File size and type limits.** Both interfaces impose limits on file size and the types of files that can be processed. Text-based formats (PDF, DOCX, TXT, markdown, CSV) are reliably supported. Complex formats with heavy visual layout, embedded spreadsheets, or non-standard encoding may not be processed accurately. When in doubt, converting to plain text or PDF before uploading reduces the risk of processing errors.

**Context window consumption.** A large document uploaded in full will consume a significant portion of the context window, leaving less capacity for conversation history and model responses. For long documents, consider whether you need the entire document or whether a relevant extract would serve the task better.

#### Pasting Content Inline

For shorter content, pasting directly into the message input is often more reliable than file upload, because you have direct control over exactly what the model receives. There is no intermediary processing step that might introduce formatting artefacts or encoding errors.

When pasting content inline, use explicit delimiters to separate the content from your task instruction. This is not a stylistic preference. It is a functional requirement. Without clear delimiters, the model may fail to distinguish between your instructions and the content it is being asked to process, particularly when the pasted content itself contains imperative language or structured text that resembles instructions.

XML-style tags are highly reliable delimiters and are well-recognised across all major models:

```
Summarise the following meeting notes. Focus on decisions made and actions assigned.
Ignore background discussion.

<meeting_notes>
[paste content here]
</meeting_notes>
```

Triple backticks work well for code, configuration files, and structured data:

````
Review the following JSON configuration for any missing required fields.
```json
[paste configuration here]
```
````

Labelled section headers work well for multiple distinct pieces of content:

```
Compare the two product descriptions below and identify the key differences 
in positioning and target audience.

--- PRODUCT A ---
[paste content here]

--- PRODUCT B ---
[paste content here]
```

#### Signalling What Matters

When the content you provide is longer than the specific information your query depends on, explicitly telling the model where to focus its attention improves retrieval reliability substantially. This is a direct countermeasure to the lost in the middle effect described in Chapter 1.

Effective signalling instructions include:

```
The most relevant section for this query is the "Risk Assessment" section 
on page four of the document. Use this as the primary basis for your response.
```

```
Focus on the financial data in the third table. The surrounding narrative 
context is background only.
```

```
The key constraint for this analysis is stated in the final paragraph 
of the attached document. Weight it accordingly.
```

These instructions do not guarantee perfect attention, but they meaningfully shift the probability distribution toward the content you have identified as critical.

#### When Inline Context Is Sufficient

Inline context and file attachment are sufficient for the majority of individual knowledge work tasks within GLBNXT Workspace. If you are working with a single document or a small set of documents, the task is a one-off or infrequent, and you do not need the context to persist across multiple sessions, Tier 1 is all you need. Adding complexity beyond this point is warranted only when the task genuinely exceeds what Tier 1 can support.


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.glbnxt.com/tutorials/guides/working-with-knowledge-and-context/tier-1-inline-context-and-file-attachment.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.