anvi'o v2.3.2, "Ocean"

We are happy to announce the new version of anvi'o: "Ocean" (The code name for this release emphasizes our attachment for this fragile environment, and all of its inhabitants. This code name was suggested by @ShaiberAlon, and won the popular vote. Another suggestion was Samurai Champloo by @tdelmont, and although we all REALLY wanted it to win, it managed to get only a single vote. Real pity).

After 127 changes in the codebase that introduced more than 1,850 lines of code and removed 500, as usual, the current version includes many bug fixes, as well as some major changes. This release note will give you a summary of most important changes.

Splitting large anvi'o profiles into easy-to-share, self-contained, and fully functional pieces

We have always been interested in providing our users with the ability to share intermediate findings their analyses produce. However, if anvi'o profiles can get too large too quickly, and it was not possible to share only a subset of a profile.

Now it is possible to create individual, self-contained anvi'o profile and contig databases for bins stored in anvi'o collections (#482). This way, instead of sharing the entire profile database and contigs database, you can share only a single genome bin from your profiles. This is done by our new program anvi-split, which creates a very comprehensive subset of your metagenome. It is so comprehensive that everything that can be done with your full data (from visualizing your data interactively to displaying or investigating SNVs) can be done with this subset, as well.

Leaner auxiliary data files

Now anvi'o uses GZIP compression on HDF5 files, which results in approximately 5 times smaller auxiliary data files (b83f747). The best part? It is that it is backwards compatible!

A small gift for miners of single-nucleotide variants

The inspection page of the anvi'o interactive interface allows you to investigate mapping results in a comprehensive way. We have been looking at our inspection pages a lot, and it was always a pain to learn important details for a given single-nucleotide variant directly from the interface.

With this release (13f1fbd), you can click on those SNVs, and enjoy your new popup window that lists everything that is relevant:

Single tear.

Storing anvi'o displays in batch mode like a pro

Have you ever had tens of bins that you would like to visualize in the anvi'o interactive interface and save what you see in a figure? We had to, and it required lots of manual work. So we implemented a new parameter (e467360): --export-svg. When you provide a file name to this parameter in your anvi-interactive call, and if you have a state called default in your profile database, you will get an SVG file! So this becomes totally possible to do:

anvi-refine -c CONTIGS.db -p PROFILE.db -C MY_COLLECTION -b MY_BIN --export-svg MY_BIN.svg

If you had read "Working with SVG files anvi'o generate" post, you also know that you can immediately convert this output into a PNG file, too:

inkscape --without-gui -f MY_BIN.svg --export-png MY_BIN.png -d 300 -D

Doing these two operations in a BASH loop allowed us to visualize all our bins in a recent project in a heart beat and create a figure like this with minimal effort:

If we can do this, think about what you can do!

While he was on this, @ozcan also fixed the alignment issue with our SVG outputs (f7cd4de) so our SVG exports center in Inkscape correctly.

More interface sweets

We know none of you use it, but we can now show samples and layer labels in phylogram view (#490) in case thats why you were not using it.

Also now you can use the shortcuts m and s to hide the settings and mouse panels (4447eec). The best part? They come back when you press them again! Magic.

Thanks to YOU

• Thanks to Roxanne Beinart (@rbeinart), now anvi'o exports gene alignments in protein clusters (#494).
• Thanks Damien Courtine (@dcourtine), anvi'o will no longer complain about the unavailability of IP addresses and stuff (#485).
• Thanks to John Eppley (@jmeppley), anvi'o will not fail to close files it opens (#487).
• When Joe Vineis (@jvineis) complained about the interface glitches when there are too many bins in a collection, @tdelmont suggested that we should tell him to not use anvi'o with too many bins. But we didn't listen to him, and fixed it (#491).
• When Joe Vineis came back and told us that anvi-summarize tasks were failing when custom HMMs are used, @tdelmont suggested that we respond to him by asking him to not use custom HMMs with anvi'o. We didn't listen to him, and fixed it, too (#484). Now you know!

Installation or upgrade

Please do not use the binaries below, and visit our installation page for details:

http://merenlab.org/2016/06/26/installation-v2/

Questions?

Come join our discussion group!

anvi'o v2.2.2, "Spring"

We are happy to announce the new version of anvi'o, "Spring" (a code name which was suggested by @ShaiberAlon, and won the majority vote).

After 300 changes in the codebase that introduced more than 9,500 lines of code and removed 3,000, the current version includes many bug fixes, as well as some fundamental changes to the repository that we are very happy about. This release note will give you a summary of most important changes.

Python 3 .. finally!

It has been a long time desire to switch our Python 2 codebase to Python 3. Well, we did it (24ff580), and finally closed the issue #343 with a piece of mind. If you are using Homebrew, the switch should be seamless. If you have been using anvi'o via a virtual environment, you will have to recreate your environment, which is a trivial task (please see the updated installation instructions if you need any guidance).

Much better anvi-profile

We completely revamped anvi'o's profiling program. Scalability-wise, this was one of the most critical bottlenecks of anvi'o. But thanks to @ozcan's efforts, it now can be parallelized, and its memory use can be adjusted depending on how much memory you have, and how quickly you want your analyses to finish. In fact, you now have (1c38887) a small info bar that tells you about your memory usage during profiling:

Adjusting the number of threads, memory use, and the frequency of disk access is controlled by three new flags: --num-threads, --queue-size, and --write-buffer-size.

Please see the updated help menu of the program anvi-profile for more information about these flags, and/or take a look at the new blog post for the technical details behind this upgrade and some usage tips:

http://merenlab.org/2017/03/07/the-new-anvio-profiler/

A revamped anvi'o interactive interface

If you have been using anvi'o for a while, you will probably like it a lot. Not only the interface now looks much more elegant, it is responsive (a960b00), has panels that can be hidden (8ae2f56), shows legends in a separate tab (dd7c6bc) and lets you order them quickly (49cfa29) or change their colors in batch mode (f5f7885), and allows you to change the background opacity of numerical layers (ce15b8a). With these changes, anvi'o interactive interface is likely to be one of the most comprehensive and flexible browser-based visualization environment.

If you are updating your anvi'o installation, please don't forget to refresh your browser window by pressing CTRL+SHIFT+R, or COMMAND+SHIFT+R so the cached browser files from the old version are updated as well.

Markdown descriptions

We often want to take notes as we work on data using the interactive interface. More often than that, we want to be able to communicate what the data is telling us to other people who will take a look at it later. We finally added a functionality to the interface so you can prepare or display descriptions in anvi'o profile databases (it supports Markdown syntax, so you can create pretty looking, readable summaries of the display):

THANKS!!!

We are very thankful for people who take their invaluable time to report bugs, or ask questions. We would like to acknowledge the following names for their direct contributions to the codebase:

• Daniel Blankenberg for adding the missing parameter for anvi-gen-variability-matrix (#457).
• Elmar Prüesse for ameliorating the hardcoded file name requirement for the centrifuge parser (#464).
• We thank John Eppley for fixing the NCBI blast / DIAMOND check (#441), and for embarrassing us (#443) (and by 'us', of course we mean 'Tom').

Installation or upgrade

Please do not use the binaries below, and visit our installation page for details:

http://merenlab.org/2016/06/26/installation-v2/

Questions?

Come join our discussion group!

anvi'o v2.1.0, "Christmas"

We are happy to finally announce a new stable release of anvi'o.

Between August and December we made more than 500 commits to the repository that resulted in ~8,000 new lines of Python and JavaScript code, and the removal of ~3,500. These changes introduced fixes for many small bugs, and new features.

We originally intended to release this before the XMAS, you know, as a "thank you" present from the anvi'o developers to the users of anvi'o, but it didn't work. But we couldn't let this stop us from using the holiday theme!

OK. Here is an incomplete list of things that worth mentioning.

If you are updating your anvi'o installation, please don't forget to refresh your browser window by pressing CTRL+SHIFT+R, or COMMAND+SHIFT+R so the cached browser files from the old version are updated as well.

A new pangenomic workflow

Until this version the anvi'o pangenomic workflow has mostly been exploiting the visualization capabilities of anvi'o. In this release, we go beyond that and offer you a much more powerful workflow with summary (#412), and protein cluster inspection (248fb14) capabilities.

An up-to-date user tutorial for our new pangenomic workflow is here:

http://merenlab.org/2016/11/08/pangenomics-v2/

We thank Jarrod J Scott for following the development very closely and testing new pangenomic features. His feedback has been very helpful to identify problems and resolve them quickly.

Anvi'o contigs dbs can has NCBI COGs

Functional annotation continues to be a pain. Although we have been very excited about EggNOG mapper, and wrote hundreds of lines of code to make things easier for anvi'o users (i.e. a driver class and a client that can import results into an anvi'o contigs database), that endeavor ended up being a failure (due to some technical issues outside of anvi'o).

To make at least a simple option available, we decided to make anvi'o able to deal with NCBI COGs. Yes, they are not being updated any longer (because why should we have nice things anyway), but they still do quite a sufficient job. Here is a short documentation that describes how can anvi'o contigs databases can be annotated with COGs:

http://merenlab.org/2016/10/25/cog-annotation/

A new, much less painful way to install anvi'o

Homebrew. Enough said!

Please see our updated installation instructions:

http://merenlab.org/2016/06/26/installation-v2/

New programs and scripts

• anvi-gen-genomes-storage: Creates a new class of anvi'o databases, "genomes storage", for pangenomics.
• anvi-display-pan: Just like anvi-interactive, but for anvi'o pan profiles.
• anvi-setup-ncbi-cogs: Sets up NCBI COGs by downloading the data from the NCBI, reformats some files, and generates search databases for NCBI's blastp or DIAMOND.
• anvi-run-ncbi-cogs: Uses your own resources to perform search, and stores results into the contigs database.
• anvi-export-functions: Exports all the functional annotations from an anvi'o contigs database so you can import them using anvi-import-functions
• anvi-export-gene-calls: Exports gene calls from an anvi'o contigs database so you can use them later with --external-gene-calls parameter when generating a new contigs database.
• anvi-delete-collection: Removes a collection from an anvi'o profile or pan database.

And here are some scripts:

• anvi-script-FASTA-to-contigs-db: For the lazy to go from a FASTA file to an anvi'o contigs database. You should NEVER use this and do it properly. But I use it all the time. It's OK. We can share the shame.
• anvi-script-get-collection-info: Cool script that tells you completion and redundancy estimates for a bin or the entire contigs database.
• anvi-script-snvs-to-interactive: Takes an anvi'o SNVs profile by anvi-gen-snv-profile, and generates an output for anvi-interactive in --manual mode. It is kinda cool.
• anvi-script-upgrade-contigs-db-v6-to-v7, anvi-script-upgrade-contigs-db-v7-to-v8, anvi-script-upgrade-profile-db-v16-to-v17: You know what they are.
• anvi-script-run-eggnog-mapper: It doesn't work. Because it calls eggnog-mapper, and then eggnog-mapper starts a server for fast search (which is awesome), and then when it is done it sends a SIGKILL, making every parent process die abruptly. So you get your results, but you die before you have a chance to store your results into the database. You can hack eggnog-mapper to not do that, but then you would have to kill every instance of egnogg-mapper manually later to clean the memory. Sigh.

Upgrading the way we deal with single-copy gene collections

We now maintain two single-copy gene collections. One from Campbell et al, which serves for the bacterial domain, and one from Rinke et al., which serves for the archaeal domain (#401). The new HMM infrastructure (#402) and the new learning module in anvi'o makes it possible to do some other interesting things going forward.

More intuitive nomenclature

We have been using "portion-covered" as a way to describe the ratio of the number of nucleotide positions that are covered by at least one read in a given contig. This in fact is what can be called 'detection', since it is exactly what it is. Well, that mistake is now fixed (#428).

Prettier interfaces

We know. You are asking yourself "how prettier can anvi'o get?". Right?! Well, thank you. As you know, besides the standard one, the anvi'o interactive interface has multiple modes for specialized operations. Such as the pan mode for pangenomics, or the manual mode for ad hoc visualization tasks. Now the interface tells you which mode you are in, because that's the right thing to do:

Also if you have been using anvi'o for a while, we hope you will appreciate what we did with the left panel.

Other notable bug fixes and improvements

• Tom finally won, and we dropped the requirement to have the name 'contig' in the A1:1 column of additional data files (#425).
• Improved default values for the interface to make everyone's life much easier (#422).
• anvi-get-sequences-for-hmm-hits can now return AA sequences with --get-aa-seqeunces flag (#400).

Features we have in mind for the next release.

• We are aware of the performance issues with anvi-profile. Our next major goal is to take care of it.
• We are also planning to leave Python 2 behind, and completely switch to Python 3.
• We are planning to complement anvi-run-ncbi-cogs with a new program, anvi-run-pfams.
• We are hoping to expand the new pangenomic workflow with carefully crafted analytical approaches that will deliver sweets from phylogenomics.
• Finally we want to focus on anvi'server, and make it ready for prime-time.

Although we have no evidence to know whether anyone is reading these release notes or not, we thank you very much for patience with us. Please visit the installation page for instructions.

Anvi'o v2.0.2

Among various bug fixes over the latest stable release, v2.0.2 has multiple important improvements thanks to Özcan's efforts. I would also like to thank Julie Meyer, Patricio Geraldo, C. Titus Brown, Mohamed Amine Chebbi, Cody Sheik, and David Needham for their help, suggestions, and reports that improved the code base.

Here are some of the notable changes for v2.0.2:

Up to 400% performance increase for the interactive interface

Now numerical layers in the anvi'o interactive interface use a single path, which decreases the number of SVG objects dramatically without any remarkable visual distortion (#381). With this improvement, the number of objects in the Infant Gut Dataset go from 59,090 to 17,984, and looks as good as ever. This change will not improve numerical layers visualized in intensity mode.

More distance metrics and linkage algorithms for hierarchical clustering operations

Anvi'o has always used euclidean distance with ward linkage for hierarchical clustering. Now these defaults can be changed in any relevant program using generic parameters --distance and --linkage (#377). Here is a list of available distance metrics,

['euclidean', 'cityblock', 'sqeuclidean', 'cosine', 'correlation', 'hamming',
'hamming', 'jaccard', 'chebyshev', 'canberra', 'braycurtis', 'yule', 'matching',
'dice', 'kulsinski', 'rogerstanimoto', 'russellrao', 'braycurtis', 'yule',
'matching', 'dice', 'kulsinski', 'rogerstanimoto', 'russellrao', 'sokalmichener',
'sokalsneath', 'minkowski']

and linkage methods,

['single', 'complete', 'average', 'weighted', 'centroid', 'median', 'ward']

Import / export functionality for anvi'o interactive interface states

Anvi'o interactive interface states are files that hold visual settings of a project, and are stored in the profile databases. Now these states can be exported, and imported with two new programs, anvi-export-state and anvi-import-state. This way visual attributes can be edited manually or programmatically (#382).

Better documentation for importing external gene calls

So you will less likely get lost while trying to import your own gene calls while generating an anvi'o contigs database if you don't want to go with Prodigal.

anvi-summarize works again!

Cody and Needham found out (link) that there was a bug affecting anvi-summarize when external binning results are imported into an anvi'o profile database. The bug is no more (#380).

Setting taxonomic levels to display

All interfaces focused on species-level taxonomy when taxonomical annotations were available in an anvi'o contigs database. We have now added a new generic parameter, --taxonomic-level, so you can specify which taxonomic level you would like to focus on in appropriate interfaces (i.e., while displaying a project or when summarizing a collection).

Anvi'o appeared in the second ECOGEO Workshop on Environmental 'Omics as a platform to benchmark various approaches for metagenomic binning.

The anvi'o tutorial I used for this workshop will soon be available online.

Thank you for your interest!

---

Please don't install the new version using the archive files below. Instead, try one of these:

Installing from source | Installing via PyPi | OS X installer | Docker image

Anvi'o v2.0.0

6 months and 955 commits later anvi'o v2 is ready for prime time!

We fixed many bugs, and we further solidified the foundation of the platform. It would be nearly impossible to list all changes, hence in this late release note I will only mention things that I found important.

A new pangenomic workflow

Which requires you to install almost nothing, and allows you to combine metagenomic bins you identified from your samples with genomes from elsewhere seamlessly with the new program anvi-pan-genome. Please see the updated tutorial:

http://merenlab.org/2015/11/14/pangenomics/

An anvi'o server for data visualization

Now you can use anvi'o to visualize any tabular data, organize its layers, add normalizations, set colors, save your work, and share it with others publicly or privately through http://anvi-server.org.

Please see the documentation of the anvi'o interactive interface and data types it works with:

http://merenlab.org/2016/02/27/the-anvio-interactive-interface/

Here is a tiny demo for you:

http://anvi-server.org/public/meren/interface_demo_VII

And here is a hint for geeks: anvi-push -h

We thank Tobi Paczian for anvi-server, who temporarily joined us, and worked relentlessly and exclusively on the server interface.

A new mode for anvi'o interactive interface: "collection"

Until now there were three modes to run the interactive interface: full mode, refine mode, and manual mode. With the new collection mode, you can use the anvi'o interactive interface to show metagenomic bins in a given collection and their distribution across your samples instead of individual contigs.

More here: #359

A more comprehensive infrastructure to characterize single-nucleotide variants

Anvi'o does a much better job at profiling and reporting SNVs, and generating very comprehensive outputs through which you can query SNVs with extensive control. A tutorial is here: http://merenlab.org/2015/07/20/analyzing-variability/

Related issues addressed: #274, #364, #367, #324, #309

Better utilization of external gene calls

Anvi'o now uses Prodigal by default to identify open reading frames in contigs upon the generation of the contigs database. Which allows us to do much more with open reading frames and gene calls. But we also improved making room for the user to import their own gene calls. See anvi-gen-contigs-database -h.

Related issues: #368, #290

Better documentation

Please see the updated anvi'o project page with new and updated tutorials:

http://merenlab.org/software/anvio/

A new sticker for anvi'o

Yes! It is true:

Come say hi to us in Montreal during the ISME conference, and we will secretly put one in your pocket.

---

Please don't install the new version using the archive files below. Instead, try one of these:

Installing from source | Installing via PyPi | OS X installer | Docker image

Version 1.2.3

Although we had no plans to make a release from the v1 branch after v1.2.2, there have been some bug fixes in this rather stable anvi'o branch, so here it is.

Thanks for your interest!

---

Please don't install the new version using the archive files below. Instead, try one of these:

Installing from source | Installing via PyPi | OS X installer | Docker image

Version 1.2.2!

This release

After this release the anvi'o master repository will carry on with the 2.* versions, which has fundamental differences from the current implementation at how genes are identified and processed, and version 1.* will be kept in a separate branch. Details about 2.* will be on the blog in the near future, but here we are with 1.2.2! Please don't forget to update your installations, because among many small fixes, with this version we have,

• An anvi-refine that works with single profiles as well (#227).
• A new flag for anvi-summary, --quick-summary, which generates a quicker summary just to see an update of bins during the refinement of large collections (a538187).
• Also, positions of samples do not change in the interface as a group when sample ordering is changed from the Samples tab (fdd8439).
• Plus, due to continuous requests, newly created bins now have random color assignments again (e572818).

Anvi'o and pangenomes

In case you missed it, we recently published a blog post on how to visualize pangenomes using anvi'o. This release also includes a key script for that workflow, anvi-script-itep-to-data-txt, if you decide to use ITEP to generate your protein clusters. Here is a screenshot of anvi'o's take from a simple pangenome:

Thanks for your interest!

---

Please don't install the new version using the archive files below. Instead, try one of these:

Installing from source | Installing via PyPi | OS X installer | Docker image

1.* clade is coming to an end

We are happy to release the anvi'o version v1.2.1, most probably one of the very last versions of 1.0 clade, as we are working on a major change in anvi'o under #121.

This release contains multiple bug fixes and improvements, including:

• A fix (387200c) for a bug we introduced with version v1.2.0, which probably ruined a lot of first-time-users' experience with the interactive interface since the log-normalizations were not done by default. We of course blame Tom for that bug, and apologize for any inconvenience it has caused.
• Now anvi-interactive can be invoked with a state id parameter (#211): look for --state flag in the anvi-interactive -h output.
• Now anvi'o computes prettier labels for bar charts (#222).
• Finally, anvi-matrix-to-newick program has a new flag: --transpose (7137981), so it would be a bit more useful to people who wish to use anvi'o for the visualization of pangenomes.

---

Please don't install the new version using the archive files below. Instead, try one of these:

Installing from source | Installing via PyPi | OS X installer | Docker image

Farewell, SUMMARY.cp

Apart from fixes for many tiny bugs, we have some major changes in v1.2.0!

By the way, the anvi'o paper is out, check it out: https://peerj.com/articles/1319/

Killing the SUMMARY.cp (with fire .. and HDF5) (#202)

• We have been storing coverage and variability information for each contig in a separate, serialized Python object file in the output directory. Although it was pretty useful for speed purposes, with thousands and thousands of small files, this solution created a huge overhead on all file system related operations. I.e., it was a pain to download or copy/move anvi'o analyses. With this release we're switching to a much better solution using HDF5. From now on there will only be one file, by default called "AUXILIARY-DATA.h5" in the output directory. You can generate this new AUXILIARY-DATA.h5 file from your old SUMMARY.cp files. Please see anvi-script-generate-auxiliary-data-from-summary-cp script to see how to upgrade your previous analyses (it is really very easy!).

Other changes you should know about

• We changed the way the variability data is generated (afd7a57) (which will affect the variability view in the interactive interface). Instead of the previous heuristic which summarized the single nucleotide variations in a given contig into a single number, we now use the "variation density" concept the way we defined in the anvi'o paper (which corresponds to number of reported variable nucleotide positions per kb; see Figure 3 in the publication to see how it is employed).
• Anvi'o interactive interface can be run in an ad hoc manner with user-provided files instead of anvi'o databases. However, this mode did not support the storage of state files or collections from the resulting tree. Thanks to Lois Maignien's push, we changed the way anvi-interactive works. Now it does support the storage of states and collections (#203).
• Now the taxonomy is properly reported in the summary outputs (thanks to Rika Anderson's report) (#201).
• anvi-profile -i BAM_FILE.bam --list-contigs works again (thanks to Linda Amaral Zettler's report)(#200).
• Now we have a TAB-delimited matrix for the overall summary of bins, and there is a download link for it in the summary output. Here how it looks in the infamous mini_test (#204):

---

Please don't install the new version using the archive files below. Instead, try one of these:

Installing from source | Installing via PyPi | OS X installer | Docker image

Samples Information database and more

Apart from fixes for many tiny bugs, we have some major changes in v1.1.0.

A new anvi'o database: 'Samples Information'

• The purpose of this new database is to store information about samples (#110). Initially it leverages anvi'o's ability to do more powerful visualizations, but in the long run it is going to be the basis of helpful statistical analyses. There will be a blog post on http://merenlab.org/projects/anvio/ about the basic usage of this feature. Here is how this feature looks like on the infant gut dataset with a mock samples information data (also please note the new title (3c7935e), and neatly organized labels for categorical values underneath the tree):

Other important changes

• We stopped using the word "metadata" in the codebase (see #194). We now follow the "everything is data" principle, and we are less lazy and explicit about what kind of data they are.
• We had a major change in our database terminology as well: starting with this version, anvi'o's 'annotation database' will be known as 'contigs database' (see #198).
• Sandbox directory in the codebase contains ad hoc scripts that are not fancy enough to be in the anvi'o distribution. However, from now on they will be installed on your system. If you type anvi-script-, and then press TAB key twice, you can see a list of all scripts.
• anvi-gen-variability-profile has now a new parameter, —min-coverage-in-each-sample, which allows the user to define a coverage value of X in order to make sure all final variable nucleotide positions are covered at least X times in each sample (#196). You can find more information on anvi-gen-variability-profile here: http://merenlab.org/2015/07/20/analyzing-variability/
• We changed the underlying design of our command line client user interfaces (ef78ee4). These changes are largely seamless to the user, however, some parameters are not the same anymore. Please re-run your batch scripts, if you have any, to make sure you have the correct parameters after upgrading your databases.

Database upgrade homeworks for long term anvi'o users: Booo!

Yes, sorry! If you have been using anvi'o and if you are upgrading it from the previous version, there are multiple things you will need to do:

• Convert your existing 'annotation databases' into 'contigs databases' using anvi-script-convert-annotation-db-to-contigs-db, and don't forget to rename them if they are called 'ANNOTATION.db' just so you remember later.
• Upgrade your existing contigs database (which you just converted from an annotation database) from the previous version 2 to the new version 3 using anvi-script-upgrade-contigs-db-v2-to-v3.
• Upgrade your existing profile database from the previous version 5 to the new version 6 using anvi-script-upgrade-profile-db-v5-to-v6.

Anvi’o developers apologize profusely for all the inconvenience, and they promise that this was the last change at this magnitude (furthermore, they agree to toss @ozcanesen into a volcano as a sacrifice for forgiveness if they fail to keep this very promise).

---

Please don't install the new version using the archive files below. Instead, try one of these:

Installing from source | Installing via PyPi | OS X installer | Docker image